How do we mitigate some of the unintended consequences of AI, to security, productivity, and the overall job market? And, importantly, can the industry keep developing at speed and scale while also moving forward responsibly?

The program, presented in association with Accenture, will feature intimate one-on-one conversations, engaging panel discussions, and lively roundtable sessions.

Brainstorm AI Singapore 2024

Ritu NARAYAN, Founder and CEO, Zum Moderator: Matt HEIMER, FORTUNE


Brainstorm AI Singapore 2024: How AI Is Making The School Bus Run Smarter, Safer, And Greener

Conferences

Thank you all for being here and we're so glad Richard that you're here with us. Um I wanna frame things here with a little uh personal story. Uh I have a cousin who's in his early sixties and he drives a school bus in the suburbs of Minneapolis and he swears up and down that the bus that he's driving kids to school uh in every morning and every afternoon uh is the very same school bus that he was riding in in 1974 which is to say, I mean, his initials are carved into the back seat, the whole works, which is all to say that this is an industry. Uh as you might say, school bus transportation is an industry that is ripe for disruption and that is what uh Ritu and Zoom uh have been doing. Uh So tell us a little bit more about how, how you chose this business challenge. You know what, why, what made you decide this is a problem that was urgent to tackle. Yes, thank you, Matt. Excited to be here in Singapore. So Zoom uh also a student transportation problem that has not changed in 80 years. Just you would, as you were describing kids today, pretty much are riding the same way they were riding for the last 80 years or you and I did our grandparents did and a few years back, I was working for ebay when my kids transitioned to school and I was seeing every industry around me getting disrupted, using technology. But when it came to my personal life or picking up and dropping off my Children, uh it was the dark ages and interestingly, 30 years before that my mom who was an educator had left the job for exact same reason. And it was kind of a curiosity of why nothing had changed in this area. So with zoom, what we have done is we have reimagined the entire student transportation by connecting every single stakeholders from parents to Children, to drivers, to operations to schools on one single platform to bring transparency, efficiency, safety, and reliability. And so so to make this work, the the the the Children have an RF ID card. Is that right? That, that tracks their location and it's tracking through a network that uh brings the buses and the schools together. Yes, absolutely. So our routing technology has all the routes in in the system. Drivers are using tablets and when the child is on boarding the bus, they tap to RF ID card. Parent gets notified your child is on boarded, the bus school knows where your Children are and when they get down the bus, they tap the card again and uh they, everybody knows in the system that they have off boarded the bus. It also we have also been able to keep timeliness by doing that by providing this the system across the country. We have 98% plus on time uh arrival and transportation for Children, which matters a lot. Yeah. Yeah, this is, this is good. It's better for attendance, right. It's better keeps the, essentially it improves the education experience for the child and for parents, especially, uh, parents in the service economy with, uh, jobs with the schedules aren't flexible. They don't have to worry about missing a day of work if the kid can't go to school. Absolutely. I mean, you wouldn't believe it's such an antiquated system using paper and pen and, uh, tracking that way that parents used to tell us I can track my package using Amazon. I can track my pizza but I have no idea where my Children are and why is that? So, and that's the problem we are solving because this is an A I centric crowd and it is fortune brainstorm. A I talk a little bit about the role that A I plays in this kind of tech stack that you're bringing to the schools and the parents. Uh, absolutely. A I is central to a business and I'm going to talk about three specific use cases which might be very new to this particular group. The three specific use cases where we use A I is number one is our A I driven routing technology and parent app. The number two is uh is our vehicle to Great Charging program, which I'm going to talk about a little bit. And third is the operational productivity given how many people are involved, thousands of drivers and operational staff, we how we use it. So let me talk about each one of these in a little bit more detail. The first routing technology, what it does very soon in a couple of weeks uh us has a back to school. Hundreds of thousands of students are going to go back and we have to design and redesign the routes for every single school district according to the local conditions. And our routing software is able to create these routes very quickly in record time. What would take months for these school districts and in real time optimize them based on various variable like traffic conditions, the school schedules that are there, the preferences and classes of Children that are happening in different variety of school. And by doing that, we are able to optimize the commute time for students and that is huge. And if heaven forbid a bus were to break down the system could very quickly adapt to that and reroute and scramble to make sure everybody got where they needed to be. Absolutely. And that's, that's the kind of peace of mind. To everybody in the system that no matter what this technology will really help guide and use the existing infrastructure to continue to make the transportation happen every single day. And our parent app basically provides a very accurate eta parents can track as you mentioned, the kids in the real time and they, they know exactly when the bus is arriving when the bus is leaving. And I'm sure a lot of us are parents in this audience and myself as a parent, it is frustrating when you don't know where the bus is. So it brings not only peace of mind to parents and all, all the schools and school districts, but it also reduces the commute time for students and save school money by design the system in such a way that you are using the most optimal size of the vehicle. And the money saving is an urgent point too for public schools in the United States which are very often quite under-resourced. Um Talk about vehicle to grid. That's um many of the buses in the fleet that you're introducing are electric. Uh How does the V two G system work? What's it designed to do? Yeah, absolutely. I would love to share with you that electric school buses are having a moment. And the reason for this is electric school buses are school buses are the ideal asset to be electrified. They are largest battery on wheels typically 4 to 6 times of Tesla battery. And they are one of those uh transportation vehicles that have very unique commute pattern commute locally and they're not used for transportation in the peak demand of energy, which is typically evenings and summertime. So what we have done, we have made a decision right from the day one, that all of our electric buses are equipped with bidirectional charging. When the buses finish the transportation. At the end of the day, they come back to the loft, they get plugged into the grid and using our A I driven vehicle, the grid technology, they provide energy back to the grid. And by doing that, we, they actually create a very powerful virtual power plant which is powering the homes in, in a very clean energy way in the local communities. So it's such a big win, win, win. It not only reduces the emissions, the electric vehicle, but it's also reduces emissions and the grid resiliency in the community. And A I is overseeing a lot of that interchange to make sure that the power needs match the Absolutely. It's the A I driven virtual power plant which knows exactly when to charge the bus buses, when to discharge them. And what's the optimal way to do it? Got it because uh I I warned you earlier that our time would fly and our time has flown. But I'd love to know. Uh Ellie mentioned that you've recently had uh a new funding round that among other things gave your company unicorn status. So, congratulations. Um What are some of the plans? How do you plan to deploy the new financing? What are the next steps uh in the building of the business? Yeah, absolutely. Uh I'm proud to say our current round. Last round was led by GSE which is a Singapore based sovereign Fund, but invest globally across the world and they look for businesses that are, that are scale can be scaled and very long term in outlook. So we are very proud to partner with them in A I using the A I to decarbonize the transportation industry and make a big impact on society. And if I were to talk about our plans, it's the momentum, essentially momentum and expansion and momentum and electrification. We are in 4000 schools across the country in 14 states and we recently announced Oakland is the first uh school district in the country to be 100% electric with bidirectional charging, giving 2.1 gigawatt hours of energy back to the grid. So our goal is we are not going to stop until every child across the globe has that access. That's really remarkable, rich Narayan. We're so grateful, so grateful that you're here. And so I wish you all the best of luck with Zoom. Uh Please thank her for her time. It's so great to have her here. Such an honor. Thank you.

Khairul ANWAR, CEO and Founder, Pandai Tim BALDWIN, Provost, Mohamed bin Zayed University of Artificial Intelligence Sonita JEYAPATHY, Co-Director, Center for Pro Bono & Clinical Legal Education, National University of Singapore Moderator: Jeremy Kahn, FORTUNE

Brainstorm AI Singapore 2024: AI in education—fighting cheating and preparing smarter students

Khairul ANWAR,  Pandai, Tim BALDWIN, Mohamed bin Zayed University of Artificial Intelligence, Sonita JEYAPATHY, National University of Singapore, Jeremy Kahn, FORTUNE

Sunita, let's start with you. Um at, at the National University of Singapore, your law school. How are you finding students are using A I? And how are you preventing them from misusing the technology? Why you need to do certain things? Why you need to learn certain critical skills with or without the assistance uh That all helps to solidify for the student whether or not to use this particular tool. So yes, there will be students who will, who if they are not better aware, they will try to use it, but sometimes that's not really their fault as well. They need to be made to understand why certain tools should be used perhaps at a later time after they've acquired certain critical skills without the tools. And do you have a specific policy about use of chat bots? Um So we uh there is anti plagiarism obviously uh in in any tertiary institution. And the knee jerk reaction when Chat GPT came out was um what are we going to as educators? What are we gonna do? Are all our students gonna be plagiarizing things from Chat G BT? Um But then after a while things settled down a little bit more and we realized that we could leverage it as well. I mean, there was a lot of experimentation but it could be a tool that could assist students, it could be a tool that could assist educators as well. So we're still in the throes of experimentation, but students basically know they can use it, they need to be aware of plagiarism policies and they need to also take personal responsibility for how they use it. Right? And when in doubt, talk to the educators, right, think about what the institution allows you to do. Absolutely, Cairo, you've created a kind of a I tutor. Can you talk a little bit of how this works and how do you make sure the tutor is providing information in a way that the student is actually learning as opposed to sort of enabling them to sort of shortcut through the learning process and maybe not actually take on board the information that they should be acquiring. All right. All right. So just a little bit of context. Panda is an education app to help, you know, school students outside of school or after school to do, you know, revision study to access new educational materials. And Panda started in 2020. Now, we have more than half a million students using it. So um we incorporated I and one of the features is a chatbot that helps students with their homework, with their assignments and all that. But what we've done different is that as unlike chat G BT, the chat board in Panai actually accepts, you know, students lingo, it is trained on the curriculum. And what's important is that it's designed not to give the outright, but you have to give you step by step solutions. So um to ask the students, you know themselves, what's next and what do you understand now? And where do you think the next step is? And it's as if a personal tutor is sitting next to the student asking the students questions rather than, you know, giving the answer outright as if it's a calculator. Right? So the danger is that with, with tools like JA I and CT like 25 mission is that students would outsource everything to the A I, right? And we've heard many times that, you know, um we can treat JN I as another education tool like cos or when two came out or Excel or sheet, we can move on to more complex stuff. But the danger is that if we allow that to happen, the student will outsource not only their writing but also the thinking process behind it. Right. So with calculator, the student have still think about the question, what numbers to input, what pressure to put similar with Excel shape and Google, etcetera. But with gen I, you can just copy everything into gen without even reading it and then you get the answers. Right. Right. Which is quite presentable. So that's the, you off the thinking. So, right. And then you use a sort of socratic method to prevent, you know, students from sort of leaping ahead to just having to do everything for them. Tim. I, I know it. Uh mmbz you've been using LL MS to help, uh you've created LLM that can help people improve their modern standard Arabic. I don't know if you can talk a little bit about that. And again, how, how do you make sure that it actually functions in this kind of tutoring role? Uh So as, as complex Mezua, I the Mohammed Bin Day at the University of Artificial Intelligence, we, we're based in Abu Dhabi. Uh Greetings from UE. Um uh We're watching very carefully what's happening in Singapore learning a lot of lessons. Uh So we're the world's first uh A I university or a graduate university masters and phd students, about 70 faculty. Um We're uh um unlike most uh universities, we're actually in the game of commercially relevant LLM model training. Uh So if you've done anything involving uh Arabic, if you've done anything involving Indonesian, uh if you've done anything involving A I, safety chances are you've, you've used uh LLM or you've used resources that have been developed at the university. Um So we, we very much sort of straddling the, the commercial academic divide uh as a university, a New Paradise University, if you like coming back to the actual question. Um uh um the use cases for, for LL MS and A I more generally in education are very real um uh in terms of research, you know, I think there's a, there's a bit of a, um, a, a tip of the iceberg effect that we're seeing in terms of G GP T. Uh And it's important to realize that there's a lot more that's happening in, in A I uh beyond just uh LL MS. Uh So, um we're also working in the space of personalized uh learning uh trying to increase both the, the access and the scalability of uh education. Um in, in part using LL MS. Yes. Uh but um very much trying to uh learn how a student is learning, uh um try to learn what are the points of confusion uh in terms of their learning process and, and cater the presentation of the curriculum and the generation of questions to uh what their weaknesses are and to try to complement their strengths uh um to uh um both increase the sort of personalized tutoring access. But if you come from a more wealthy background, you, you can, you perhaps assume uh to students throughout the world, not just in English uh in languages uh beyond English as well. And I think that it's very important that we don't fixate too much. Yes. As educators, we, we sort of think about plagiarism and, and through cheating, it's a very real uh issue uh as a, an educator, uh that I do think about. Uh, but there are, there are so many benefits, so many good things that can come out of the A I models. Well, beyond the, the sort of simplistic chat QVD, use cases that I think we need to be aware of. Yeah, absolutely. That's interesting. I want to come to questions from the audience in a little bit. So please think of your questions and I'll come to you in a minute. Um Cairo, I know that, you know, with panda, it's not just about their supporting the student, but you've also built uh aspects of features of the product that support teachers. Can you talk about that? Because I think that's sometimes overlooked. There's so much focus on what is the student doing and, and less on. Well, how can this technology actually support teachers? All right. So when we started in 2020 our product was designed to help students directly. So students will go to school like they normally do. But after school outside of school, instead of using paper based stuff they can use instead. And then um around a year and a half ago, teachers actually came up to us saying, hey, um is there something we can do for them? Right. So now we have a new module called Panai teacher where students can actually, teachers can actually teach there have their own ID and have special functions. That's what that for them. So I don't want to get into so much detail, but it it covers, you know, teaching, teaching planner, it covers monitoring the students, analysis of the students as well as self development content to develop themselves the teachers themselves rather than just to develop the students. So what we've seen is that there's um you know, a valid concern on the over reliance on A I interaction that diminish the vital interaction with the students and their teachers, right? So um it's it's important that no any tool that we develop actually enhances the interaction with teachers as well. And that's why, for example, in pan, we have the par model, the teachers model try to complete the ecosystem somewhat and, and and we've seen even with our chatbot when students are, um they um initially, yes, they were asking academic questions, but after months of, you know, using the product, students actually started to share about their personal life, sharing about their parents, sharing about their boyfriend, their friends, about the problems in school, their friends, right? And confiding into the chat board, right? So that's also a danger where you know, young students, especially because we were dealing with K 12 primary school students as well, treating the board as if like a friend, right? So replacing the vital interaction with his students and their friends and social interaction with teachers, et cetera. So that's the den. So um the, you know, the ethics behind it so interesting. I, maybe we'll talk a little bit more about that. This sort of the way these, yeah, maybe this, this morse from being a tutor to kind of being a companion chap bot and, and maybe there's a sort of tendency to, to go that way. Um, uh, so I wanted to ask you if you, do you feel like, uh, either the way teachers have to teach, needs to change or what they have to teach needs to change. Uh Because of gender of A I and because of the, you know, the popularity of these chef bots, um So I'll share the perspective, the legal perspective since I'm with the law faculty. So I guess projecting into the future, the now, maybe we know that human and machine will need to work together, right? There needs to be some interact, there needs to be interaction there and how the unit operates, leveraging the strength of each that will make it a AAA positive um combination. We also know that things will change extremely quickly and people will need to learn and unlearn skills and knowledge extremely quickly. So, taking that in mind and coming back to the curriculum that we teach at the law faculty, the question then is what do we need to change? And how might we need to change how students learn things? Obviously, there will need to be a law and tech element. Um And by this, this exposure to technology can come in various different forms. Students need to be aware of obviously the latest digital products out there so that they can think about legal issues that can arise from that there can be exposure in terms of how educational tools personalized learning because then they can see to what extent it is useful or not and the strengths and weaknesses of the tool. So playing around with that, taking personal responsibility around that um exposure to what's happening in the industry because that changes at a breakneck speed as well and learning how to deal with um or do the digital speak, I suppose in the in the foreseeable future where um students or lawyers need to speak to technologists, right? To build products that could help the industry. Um But apart from that, I think in terms of just thinking skills, um you want to have higher order thinking skills like analytical skills where students can look at complicated things, break it down into its component parts. Understand the relationship between the two, you need synthesis skills where they can look at different threads of thought and create new patterns. You need also the values that students are invited with because they need to understand why they are doing things. They need to understand why they are serving the community and those things aren't changing. Those are things are sort of always been there and they remain consistent, right? That's, but I think now they're more important more important than ever. Yeah. And if you don't kind of celebrate the humanness of it, I mean, that's what your combination and the human in the loop is meant to be. I want to get questions. I know there's at least one question from the audience. So if we can go to the gentleman up here in front, he had his hand up and then I also see that someone in the back. Um, but the gentleman up here, I saw him first. So, uh I don't know if we can get a mic to him, uh quickly, if not just stand up and, and say your name and where you're from and, and what your question is, should know how to use A I, whether it's the earlier predicted A I and A I to be ready for the A student has to learn that otherwise there could be universities which are doing it better than you and your students get left behind. So rather than saying plagiarism or let them use Google came in and people started using Google. So the whole curriculum has to change in terms of asking questions which are about critical thinking where they just can't copy paste from chat G BT, right? So, uh a student has to be well versed with all the latest stuff uh and change the curriculum, change the question and you're seeing very rightly that students should be able to work with technology. A law student should be able to work with technologies to come up with new A I models which help uh law. That's a good. Do you have a, do you have a question there? Is it just sort of a comment on, on that? I'm saying that the critical, the questions have to change in the curriculum has to change. I think that's echoing what Sunita was saying. Which is that, yeah, that maybe you change a little bit to emphasize the these skills even more. Um now because you have technology that can do a lot of the, the simpler things. That's what you were saying, Tim, do you, do you agree with that? I mean, I do and I think this is something that all universities are grappling with at the moment. I think even in the A I space for us as a university, our focus has been very much on education of, you know, the, the A I innovators, the people who are going to be driving the next wave of uh um A I innovation for in research. And absolutely, we need those sorts of people. Um But one question we've been asking ourselves is what would an undergraduate curriculum look like if we had AAA blank slate uh based on the existing technologies that are already here as well as what we know is coming? Uh What should that look like? Uh And the answer isn't necessarily that, you know, they need to understand linear algebra and they need to be able to, to uh train in Pytorch. Uh It's possibly there's an even greater need for A I super users. People who understand how these different models work. They understand the weaknesses of the models, they understand why there are going to be shortcomings in terms of um you know, whatever there was a lot of talk about through culture values uh um in city answer. But why is it that you get biases in models? Uh How is it? I can use the most successful, how can I buy them in different settings, whether it's uh legal settings or, or business settings? Um But not necessarily the, the sort of core skills we able to train LL MS uh you know, they, they, we don't need the world to be awash with LLM trainers. You get a couple of 1000 of those in the world and you're set really. Uh there, there's a lot more innovation to be had but uh really, it's the flow through effect. Uh and the, the middleware if you like and, and I think uh for a modern A I curriculum, even in computer science, I think these are the things we should be thinking about rather than, you know, the sort of classic, teach them algorithms and data structures. And, you know, have you taught the way I was taught as an undergraduate? No, I don't think that's the case at all. Interesting. I know there's one other question uh the woman has her hand up uh very high in the middle of that there if you Yeah, yes. If you can stand up and we'll get your mic to you very quickly. And if you could just say who you are and what your question is. OK, thank you so much. Uh This is one way from Singapore and really happy to be here to hear your perspective because like the definition of education has like widened with like technological cycles, getting smaller and smaller. So my one question for everyone here is that uh how do you, what is one practical tip that means not theory, practical tip that you have for everyone here because we are all here to learn, to decide what to outsource and what not to outsource. Yeah. Keen to learn from your 11 practical tip Cairo, they have one practical tip for what students can. So maybe rely on an A I tutor for and what they shouldn't rely on an A I tutor for II I think um on the student side, they definitely will try to cheat, right? Ie the fastest, the most efficient way to cheat, right? But um like this gentleman has said, um teachers have tried to change the way they do evaluation, right? For example, like um um now uh exams are being done on paper, essays are submitted with editing history now and then assignments are increasingly more personalized. So to force the students to answer and include their personal stories as well, right? Rather than something they can copy of the internet only. So I just want to highlight that it's important to highlight to the students. What is education in the first place? Right? If, if um the message is is we say you go to school, you go to college, you buy a big house, you buy an expensive car. Then if that's the case, the message is that this is just a transaction and obviously students are going to cheat. But on the other hand, if you say it is actually a way for us to for students to become an informed citizen, a better decision maker better, then maybe that's, you know, the way it is supposed to be. So we can ask for philosophical question to our kids, but it invites us to, you know, rethink how we think about education as well. So the mental of education. So I need to do you have one practical sort of tip or or thing that's about output what you should. So I second I rule with what it means to be human. Like I said, the human and machine interaction, right? So you need to leverage on your strengths and what are those strengths that will be important? Your critical thinking skills, your interpersonal skills, your emotional intelligence skills, I think you make sure you're not out sourcing those skills, right? Because you're just cheating your future self essentially and Tim, you have one tip for people. One tip agency and accountability are the two things that we have to hold on to. So whatever you're submitting as a student, it is your work. If you're using A I tools as part of that, you need to own it. So if it's code, say you have to be able to tell me what every line of code does that you're submitting, even if you've used code uh to generate it. Uh And you need to own it. And I think it's important to realize that a lot of these issues we're talking about aren't actually new issues. It's, it's a new guys uh because we're talking about A I tools. But, you know, contract, if a rising has been around, whether you like it or not for a number of years, we've been grappling with it as universities. The big difference here, as I, as I say is agency, I can use these tools to uh um enhance the work that I'm doing and embed them into my work flows. But I am ultimately the owner of what comes out of it and I have to be able to defend everything that's there. If I'm doing uh getting someone else to write my essay, obviously, II, I can't do that. This is me just with uh you know what, whatever uh um uh embellishments using A I tools. Great. Well, that's a great note to end it on. Um we're out of time. I'm afraid if we could go on a lot longer. But I want to thank my panelists and thank you all for listening. Thank you.

Panos MADAMOPOULUS-MORARIS, Founding Managing Director of Industry Programs & Partnerships, Human-Centered AI (HAI) institute, Stanford University, Anant MAHESHWARI, President and CEO, Global High Growth Regions, Honeywell Moderator: Clay CHANDLER, FORTUNE

Brainstorm AI Singapore 2024: Growth, Wages, Jobs, And Wealth—Debating AI’s Economic Impact

Panos MADAMOPOULUS-MORARIS, Human-Centered AI (HAI) institute, Stanford University, Anant MAHESHWARI, Global High Growth Regions, Honeywell, Clay CHANDLER, FORTUNE

So there has been um tremendous rhetoric about the advent of A I and what it means to firms, what it means to the global uh economy. And we have heard a huge range of uh you know, from techno optimism to techno doom is uh about, about a is impact. Uh you know, people comparing it to the industrial revolution, electricity. I've even heard people compare it to the discovery of fire, which feels a little over the top. Um And we've seen a number of really uh fascinating studies come out. I mean, uh the, uh you know, there was one that mit economist Darren uh as molo uh came out that was quite pessimistic. Uh mckenzie's had one that was a little more in the middle. Goldman Sachs has had one that uh has estimated that it could add as much as 15% to the global economy uh over the next 10 years, which seems extraordinary. Um So I'd love to, you know, think about those a little bit with you. And I'm, I'm just reminded of uh the seventies and the eighties when we had all these massive investments in, in uh computer technologies and robotics. And there was this great conundrum among economists because as, as, uh you know, Nobel Prize winner, Robert Solo observed in 1987 that you could see evidence of the computer age everywhere, except in the productivity statistics. It wasn't really showing up. Well, eventually it did show up. But I, I wonder whether, you know, they called that moment the productivity paradox. We're seeing a similar productivity paradox uh with A I and both of you are kind of perfect people to address this topic. Uh You know, an not you're right on the front lines of this uh at Honeywell, tell us a little bit about what you think about these kind of wildly disparate macro productivity uh estimates. Yeah. And maybe just to start optimism and pessimism can be seen both ways depending on which side you're looking at it from. But just a quick context of where I come from, we operate in three mega trends on the planet. That is the future of aviation, automation and the energy transition. So my comments will be centered around that. But these are real world industries. These are, these are really people who operate aircraft, people who are operating buildings and plants, manufacturing plants and people who are in the in the energy business. Now, we recently did a study across 1600 people, executives across different companies in 12 countries, with each of them having 1000 people at least. And we realize that nearly 17% of companies have actually implemented scale sort of projects which are related to productivity primarily because that's the number one play that people are seeing around productivity and efficiency. There is of course another play which is around cybersecurity where people are deploying a lot of the productivity I would say is really around saving cost of energy saving materials and getting more with less, say a little bit about your, your role uh at, at the Stanford Center and, and love to hear your take on this kind of, you know, paradox. Yes, thanks for having me Clay and excited to be here. Always excited to be in uh uh Southeast Asia. So yeah, at, at Stanford, the uh Institute of Human A I, we are short of the umbrella organization for all things A I encompass and we are very committed in uh uh advancing um A I policy research education and practice underscoring practice to uh accelerate human condition as part of this. Uh We work with a number of uh Fortune 50 companies uh help them accelerate the I journey and this means different things to different people. So there is a been indeed um a lot of uh skepticism around productivity or Roy, a number of reports came out uh both from uh uh analyst in Wall Street, but also in, in our world in Silicon Valley. Um But also economists are following this very closely. You mentioned productivity paradox uh uh Eric Bloo and my colleague at DRE has been doing work on this topic for the last 30 years or so. Um Let me unpack. Uh let's have a couple of things here. We should be unpacking one in terms of the concerns. Yes, there are some really credible concerns. Um People are looking at the Capex. Um This Goldman report mentioned, you know, a trillion dollars in Capex uh built out, built out infra hyper scale is alone, they are committing 60 to 80 billion this year uh in excess of the already committed Capex for uh for research. And then um in start up land, you have unlimited uh amount of capital uh getting into A I start ups and we don't have at the same time that many start ups in the 100 million territory. So people are skeptical of what extent to those start ups can live to the valuations. They uh they, they raise money and then for capital allocators, there's a lot of pressure because companies now uh stay in private for a long year. And with scaling laws advancing uh the hyper scalar and the magnificence seven MN AM N A will become harder and then it will be more challenging for them to build sustainable modes. So two things on our end, having worked with a number of those fortune 50 companies over the course of the last three years. One, I think people on more of the camps that we are um underestimating uh the long term impact of A I at scale and overestimating the short term. So that has created realistic expectations when it comes to Roy, both in terms of revenue or productivity or whatever you want, whatever way you want to measure it. Uh It's one thing to see public interest spiking and just open A I went to 100 million users in two months, but I was reading the other day when 20% of us retailers, it took them 20 years to get online. And there are multiple use cases, you know, apple revenue curve, sales force revenue curve. Many large organizations still uh they are not in cloud properly. So expectations versus reality is one like those things take time and second is what lies beyond tech. So most of the work we do with companies, it's not really on, you know, let's buy this tech or let's build this model or let's dump some A I to just replace the existing workflow. We feel that most of the value will come by thinking A I native, which means how you really re invent your business model, how you revisit your processes, uh how you upskill and rescue your people. This is where most of the value will come from, that will take more time. And I think even Goldman in this report, recognize the railroad analogy, you know, the first wave and the second wave. So we have more of this confidence Well, you said something this morning that I thought was quite insightful and a couple of other panelists on the main stage this morning have mentioned it, um which is the difference between using A I uh to achieve incremental savings uh on your, you know, the the production process that you've always used versus saying, OK, now we have to completely re invent the production process around the new technology. And we definitely saw that in the, in the computer investment waves where originally what everyone was doing was fiddling at the margins but not rethinking how they do everything. And, and, and that this long term curve that you're talking about on us where the real gains kicked in was when people finally said, oh, you know what, we got to rip up the whole legacy system and rethink it from the beginning. Are you seeing that with, with sort of the firms that you work with? And so what, what does that imply about how long the kind of payoff horizon might be for this technology? So maybe again, putting in context of very three different kinds of industries and use cases as you say, and I'd really like to do a poll in this room if you would fly an aircraft, if I said I'm not going to have the pilot in the aircraft. And, and that's the, that's the aircraft that you're on for 16 hours or eight hours and that will take you to your destination. There would be a few adventure seekers here. But uh but I would imagine a lot would say, you know, I'll wait on it a few more people do it and let the regulators approve it before I get to do this. Now, this is an industry that is short of aircraft and short of pilots in the air. Right now. Today it's regulation two pilots in the, in the aircraft to fly. You will it get down to zero? Maybe not, but can it get down to one with the power of A I and therefore have a lot more aviation going? That's possible. And that's where I think productivity and the workforce really really comes in because you can get a lot of input and truly A I becoming a copilot to, to that pilot. Now, the second industry, if I go and this is again a thing that when people are not talking about A I, they're talking about sustainability and uh and the energy transition. Now, if you look prepa to today, actually, the number of jobs have gone up in the energy industry because A I and the need for energy is accelerating the, the energy transition, there's a lot more renewable, there's a lot more uh different kind of green electron coming along and it's short of people. So when you're short of people, whether you're in the United States or Germany or Japan, then how do you, how do you augment that workforce and you again, bring A I to help you augment that, that workforce. And if you go down to the developing part of the world, they need to train up a lot of people for new industry. And automation therefore, is coming in. And that's again where you need A I to help that labor come up the learning curve very quickly. So a lot of training a lot of simulation, a lot of testing where the system might fail and therefore training somebody to do that. So, so overall, I would say there is definite use cases of productivity of workforce augmentation of training and enhancement of savings and energy that, that are, that are satisfying the CS in every company to say, you know, there is something in it, you know, it's part of the, the, the question about productivity gains is in a way two questions, I mean, it's a question about uh you know, how do we, how much do we think A I is going to be able to reduce labor costs or the cost of total production? Um But then I guess the other part of the question is what percentage of tasks or jobs do we think that A I can really be able to automate? So if you think about your airplane example, um I'm not sure I would feel comfortable getting into, I, I would get into a robo taxi but would I get into an airplane that didn't have two pilots. I might not. Right. Just because the stakes are so much higher. But if you think about what already happens in the cockpit at most airplanes already, maybe nine percent. I think people don't realize that when you get on a plane, pilots aren't touching very much until the actual, you know, the take off in the landing. Right. They're focusing on those really critical tasks that, that still haven't been automated. So I think there's probably a lot of equivalent to that in the production processes of different uh companies. Uh you know, Darren argues that mit that in fact, the number of tasks that can be automated is probably not as great as people think. And, and so that's where he gets, you know, he says probably if we get a productivity gain, it will only be 1% at 1% point a year rather than these crazy, you know, uh two and 3% kind of estimates. Um What can we talk a little bit about the, the job I impact uh of this sort of this transformation, you know, um the um you know, the famous uh William Gibson quote is that the future is already here. It's just unevenly distributed. And I think that, that there is a little bit of that associated with, with A I is that um uh the the fruits, the benefits productivity gains of this are going to be different in different parts of the economy and for different types of jobs. Uh How do you, how do you see that panel? I know you're thinking about this with your industry, uh you know, contact, we are thinking about it a lot indeed. And in multiple dimensions, um you know, there are a lot of parlors in, in A I so organizations are trying to figure out how, how do they manage in a world where alpha is so asymmetrically distributed. So there's a big delta between how quickly uh research is moving and then the time that organizations have to effectively respond and how, how they respond to this in our view is gonna define pretty much wins and losers for the next, for the next cycle. Um So there are a few things. Um I, I think this year we've seen a lot of things functionality uh from Executive suite and boards. Uh Most of the transformation journeys are led by either the CEO or the CIO um boards are very, very actively involved. As a matter of fact, I was in New York before the, I ran a poll in a room like this with uh with CIO S and I asked them, how many of you are you expecting to present in every board? And the vast majority of them told me every board we expect to go in and uh report progress on um on our journey uh On the A I side. Uh Second is, uh I think alliances and ecosystems as a as a mode uh because things are moving so quickly, there's so much internal and external pressure to leader to, to executive leaders, costs are still very, very high. Uh And human capital is here, I think that has pushed the envelope a little bit and uh expanding the surface area of creativity in terms of business models. So even organizations that were traditionally very conservative, they wanted to do everything themselves. Now you see those like global uh alliances, ecosystems and both of us spend a lot of time around the world, actually managing those ecosystems. Uh which I think bottom line is great for the organizations is great for uh the discipline like the A I discipline as a whole. You see the rise of cooperations because of this many companies at the same time, they might be competing or they might be partnering. And one last thing I would say is that uh yeah, alpha will be as symmetrically distributed at least for the foreseeable future. But leaders in their own industries can identify and curate their own pillars of alpha because Alpha has many different colors and something that we've seen again and again, in particular, the last six months is uh the human Centricity of how uh increasing awareness from executive leadership in different organizations or how they build design, deploy at scale in a human surgery way can deliver, you know, Alpha uh both internally and externally on the safety side on uh on the design side, on the implementation side. Um And I think as a matter of fact, we're gonna, um we're gonna see find ourselves uh in a few months of those transformation journeys, they will be measured. Their success will be measured by humans how and that connects to your argument about jobs uh and displacement. We got a few minutes. Uh Love to take questions from the floor if uh anyone has them. Um Yeah. Yeah, we get a, do we have four mics? Where are there? He comes? Thanks. Uh Seth Dobrin. Uh You know, I think there's three things I really didn't hear, address that are important to this conversation. One is a lack of diversity and so much. So, you know, for instance, Panos like, you know, Bee who founded, you know, the Human Centered A I is, is really the grandmother of A I and no one ever talks about that. They just talk about the fathers of A I. Um And, you know, lack of global representation um which is really important because people talk about re you know, fine tuning models, but fine tuning actually breaks the model. So you can't do that when you look at other cultures, which can cause actual type of colonialism, a new type of colonialism because you're inflicting Western and Chinese cultures on the rest of the world. Um And then the short term impacts on jobs. Yes, we're gonna create new jobs but you're gonna have a transition period where you're gonna have a lack of job creation, you're gonna have people leaving the workforce. What are things that can be done and what is how, you know, how I found in the, in the near term, you know, what's going on through the research that, that you guys have done? Thank you. Is the question addressed to me exclusively or 1.5 minutes. So let me talk quickly on, on two of them and IV line about everything else. Yeah, I did is extremely uh committed in advising uh diversity. As a matter of fact, we had our five years anniversary uh a few weeks ago and we in our dinner in our dinner session, we have the godmothers of, of, of A I and I, as everyone go on youtube uh and what the conversation was illuminating. Um We are very actively on the research side trying to empower and enable this on campus for our grant programs. Um But there are a big issues that they are affecting diversity in A I which is the availability of resources overall. It has become extremely uh expensive to run major A I experiments on campus. So, um you know, academia is losing talent to the industry. Uh And we are committed both in terms of our public policy efforts, but also our work with the industry uh to reverse that as much as possible. Uh really quick on, on on the impact of jobs um short term versus short term. Yes. Right now, we see certain use cases um that they prove that this is that the human and the machine are better where human and machines collaborate are better for organizations and better for society. Um We see that in customer support, we see that in diagnosis, there are a number of use cases and there are few papers coming out of Stanford mit other places that they prove that this, this uh there's this economic concept of complementarities where there are certain tasks that humans are better. Um emotional, emotional intelligence, creative problem, problem solving, contextual understanding. And there are other tasks like the ones that are predominantly uh repetitive in the nature and heavy on data that A is are more effective. So bringing those two together um we believe will amplify human creativity and um augment like we're very committed and physical augmentation versus replacement. Se second, talking about pilots is uh uh another economic principle we did the devis paradox. So when the Z engines were invented and technology dropped, uh increased productivity dropped the cost, the demand for pilots skyrocketed, it didn't went down. There was a lot of talk about um um you know, in the health care community uh about radiologists and how radiologists are gonna be the space. Well, guess what uh us cannot have enough. There are two X number of postings on the people who are actually available to work as radiologists. And that goes down to this physic where for some tasks, we're gonna still need humans in the loop. And we have agency, business leaders and policy makers, they have agency in making those decisions. Those are not decisions that will come embedded with products, but humans have agents in directing the strategic direction of the organization. We are unfortunately out of time. Uh But a lot of challenges, a lot to think about. I would encourage everyone to uh find both uh an A and uh and panels, uh you know, uh in the, in the hallway, in the lobby to follow up and, and uh explore more about these topics. Gentlemen, thank you very much. Thank you so much.

Brian COX, Director of Gameplay Engineering, Inworld AI Dian GUAN, Co-founder and General Manager, Asia Pacific, Patsnap Donn KOH, Co-founder and Design Partner, Stuck Design Hanyi LEE, Founder & Creative Partner, Anak Moderator: Ellie AUSTIN, FORTUNE

Brainstorm AI Singapore 2024: How AI will impact human creativity

Brian COX, Inworld AI, Dian GUAN, Asia Pacific, Patsnap, Donn KOH, Stuck Design Hanyi LEE, Anak Moderator: Ellie AUSTIN, FORTUNE

Hello, everyone. Welcome to the stage. Thank you so much for joining us. So this is the only session um at the conference that we're dedicating to the creative industry. So there's a lot to discuss and Brian, I want to start with you. It's a really simple question. Can you explain the most significant way in which A I is revolutionizing gaming? Uh Sure. So at Enro A I, we provide A I models and microservices for game developers to add all kinds of A INP CS gameplay features and content generations. OK. Um We'll come back to some of that in a second because you showed me a very cool video before we came on stage of you kind of explaining a world that you wanted to be created and A I simultaneously creating that world in a lot of detail. So, so there's a lot to discuss there. Dean that snap is a global leader in A I powered IP and R and D Innovation Intelligence. You work with companies and organizations like NASA Tesla Disney, huge organizations. How are you working with these brands and A I on a day to day basis? How is A I impacting the way that you're working with these organizations. The pa specialize in providing uh smart data and intelligence for R and D research and development professionals throughout the R and D life cycle all the way from ideation to discovery development to commercialization. Um So if you think about R and D, it's much, much more than the eureka moment of a genius guy coming out with a brilliant idea. It's actually a lot of tedious and very long process. For drug discovery example, it takes more than 10 years to develop a new drug. A lot of data collection searching, trying to synthesize the data fighting patterns and eventually come to the product that works. So pass snap build our own large language model and also a lot of small models trained on the special data set patents, academic journals and papers. So our model is like a very specialized GP T we understand R and D professional much better. OK. Dawn Let's come on to you. So ST design is a design and innovation consultancy and I believe that stuck has developed its own A I tool called hyper sketch. Why was it important to you to develop your own tool rather than use one that's already on the market for us? Uh Hyper sketch was a reaction to the A I tools out there which I think are often very much focused on trying to get the best quality of output um using the best data. But, but for us, we needed something that would amplify creativity and, and sometimes in that space, bad data actually is the best data because we are trying to get the unlikely things that surprises us and the things that haven't been good answers of yesterday. So for the main reason, we do hyper sketch as a way to amplify our work. And can you very briefly explain to us what hyper sketch does how, how you leverage it to improve your work with your clients? Oh, sure. If you take a simple doodle, half a doodle, maybe because you have half an idea with all the prompts, you could just get it to say, make some products for me and you will see in high fidelity pretty soon in a matter of seconds. So that gives you something to bounce back with like almost like a brainstorming companion. Fascinating. OK, honey, let's come on to you. So you're the founder, creative partners I mentioned of an app which is a global design company that works with the likes of Netflix Apple Bany entry. And what I'm interested in is at what stage are you using A I in the creative process when you're doing work for these massive organizations? Um I mean, we tend to break our process into three stages, the conceptualization stage, the exploratory mock stage and the execution stage for us. The middle portion of the exploratory I guess mocking stage, that's when A I has been amazing. It has variants, it has depth, it gives you, I guess a lot of opportunities and possibilities. Um But definitely not the first nor the last. And OK, so why not the first or the last, I guess? I mean, our bread and butter is ideas. Um So I guess for us if you replicate A, I, I mean, if you use A I for the first or the last, you're just going to get the instant artificial sameness and blend this, which is, I guess the worry that most people have with A I in general. And yet if you use it in step two, what does it enable you to do? And how does it accelerate your work potentially? I mean, if you're like a two man design shop, you can pretend that you're 100 men strong or you can pretend that, you know, you know, knowledge about, I don't know B house or you know, I mean, something that I have no clue about. So it allows you to pretend to be other people. It allows you to be transported, transporters, allows you to be many places at once. So I think that's really the beauty of it, Brian. I, I want to come back to you on the subject of guard rails and, and so, you know, if you are to create, as we were discussing earlier, a whole new world in gaming, how do you make sure that that A I doesn't go rogue and maybe the characters don't start talking about things you don't want them to talk about or doing things that are dangerous and not something you'd want the players to replicate. Yes. So we allow our clients to first choose which A I model they want to use either the publicly available ones or our own. And then we also have safety filters and customers can choose which filters they want to enable or disable. So for example, if you want to have a character that can talk about alcohol or politics, then you can enable that or you can also disable that so that they stay track of the topics that you allow them to talk about. Got it. OK. And I think at the moment, your A I models are, they're all focused on English language, but I believe that that's about to change. Do you want to tell us a bit more about that? Yes, that's right. As of today, we support multiple languages. So we support English simplified Chinese, Japanese and Korean. And we got some feedback from our partners LG plus and then to make sure this is culturally relevant to give a better experience to players in those regions. And why did you, my final question on this is why did you choose those languages? First of all, there are a lot of video game players based in, in China, Japan and Korea. And that's why we wanted to focus on those markets first. And that's also the, the regions that we received the most requests from like, hey, can you please support our language? And uh yeah, very excited about it. Do you see yourself broadening out to cover European languages more in the future? Is there something that's going to keep rolling? I can't make that promise yet, but I believe we will. Ok, great. Um I do want to open up the questions at some point. So please bear that in mind. And if there's anything you'd like to ask our panelists get thinking about it, um I want to go to you, Dan and I know you don't think of yourself necessarily as a creative, but you're someone who enables creatives to do to do the work um that they want to do. What's your, we, we've spoken about some of the massive possibilities of A I but, but what's your fear as we start to integrate these models into the creative process? What's keeping you up at night? And how do you think your fellow CRE your creative here should be solving for it? Well, for our case, uh you're right that we enable creative. So R and D is a very classic form of the changing, I mean, converting human creativity into new inventions and new products that benefit of us. So from our perspective, I don't see as, as a, as a uh a problem is more on the positive side when we observe how our customers embrace A I I think first of all, our customer who are the research and developers by nature, they are curious people. So they are one among the first in the organization to embrace A I and in R and D space, we already see so many success cases already. Just take one example, maybe some audience already know um in terms of drug discovery, as example, it takes 10 years, more than a billion dollars, even a new drug. And it's a very, very tedious process. And one very important step is to identify the 3d structure of A molecule. Without that, you cannot develop any drug that can cure the disease. And this used to take a very well trained professional few months at least or even years to decode one particular molecule. But deep deep mines, Alpha Pho managed to solve this very difficult problem by using A I to analyze the patterns and recognize the patterns so that significantly transform the entire industry. So we see more and more use cases where clearly R and D people see the benefit rather than the problems of using A. But of course, one risk is once this repetitive task being done by A I because the R and D process, even though it's creative, there's a lot of non creative tasks. For example, running the experiments repetitively collecting data, analyzing the data. So about one third to a half of the employees in this uh R and D centric companies are doing those tasks once this is taken away by A I, what do these people do? It takes years of training to be a professional in any particular domain. So that's one potential risk that we have to address. And I want to continue with that subject and ask Don and Honey Don if you're, you know, a young person who wants to get into creative industries in this day and age, whether that's as a musician, a writer and artist, is it mandatory for now? Now for you to also understand generative A I coding? Because in a way, those quite technical mathematical skills are at odds with the traditional creative skills. So how do you balance those two sides? I don't think you have to technically operate it, but you need to have some literacy with which these systems behave because you want to use it to get better results than what the system is giving you and everyone has the same tools. So you've got to kind of understand how to use it. Well, to get a result that is firstly more competitive than what the system gives you directly and also better than what others are doing with it. So, so that that is the core thing. I think, I, I don't think we need all to become A I scientists or you know. Yeah. OK. You need to have literacy in it but you don't need to be executing it yourself? Yeah. Ok. Um, honey, so last year a group of authors including George RR Martin filed lawsuits against open A I, um, alleging that Chat GP T was trained on their original work without their consent. And so my question to you is what should creative companies be doing to ensure that they don't become the subject of legal action for wrongfully using original source material. I know that's a big question. Um, I think that, I mean, most of the times we've used it, it's mainly for commercial use for clients. So I think that's why like I mentioned before, it should be used mostly for exploring but in the execution stage, which is usually when I guess the usual copyright laws and I guess infringement come about, I think that's when it gets a bit tricky because I mean, if it's not created by some kind of human behind it, then there really is, you know what I mean? I guess no, no right towards that at all, John. Have you got anything you want to add to that? I mean, there's a lot of similarities, even though I think the legal debate is out there with me seeing something on the street and then accidentally using it in my design. In the end, we still have to kind of check that the final output. Like what honey has mentioned is not infringing on anything else that has already been done. So it's not very different from how we've operated innovation so far. OK. Are there any questions for any of the panelists? Um If uh Yes. Oh, fantastic. Let's go with this question here. But, and if you could just stand up when the microphone comes to you, uh I believe it's coming from here and say your name and the organization that you're from and then ask your question. Hi. My name is Lois. I'm from GOV Tech, also known as the Government technology Agency. So my question is actually um kind of thinking about this whole lawsuit thing in a different perspective. So can I interpret that the presence of such legal issues facing this A I based companies are actually an indication of valid profitability or a valid market such that it's not, it's not a bad news, right? In fact, if you look at it, it's an indication, it's a proof of viability. So, um I mean, recent news about perplexity, which is an A I company uh trying to get around this as well. Um What are your thought on on such seemingly negative news? Yeah, thank you. Who wants to take that? Um So um many of our partners and clients, they provide their own data so that solves this issue. So you can either use the existing area models and most of them will tell you what they're trained on or if you want to provide your own data set, then you can and then you can ensure that all the data that the A I is using uh is, is your own and then you can avoid lawsuits that way. OK? Anything. Y yes, you can ask a question and then there's another one that we're gonna come to you. If you just wait for the microphone, please so that we can hear you. It's just coming. Thank you. Hi, my name is Steve from Spark. So yesterday the grab I was saying that they are creating a lot of creatives through A I, right. They're personalizing it. So if they are say 1 million customers in Indonesia and each has a unique need and they read their pattern over a period of time how they are interacting with Grab, so they're able to personalize it. So now ads are getting created uh through A I. Secondly, accenture has an add on where they said, OK, let's make a creator which should appeal to people. So they picked up things which uh people like something like pets for kids and they incorporate all those things and give it to and they created ad and then they told a creative person to create an ad and then they tested both and the one created by A I got better. So how do you interpret that? How the industry, the creative industry, the TV ad industry is going to shape up with a coming into picture. Do you want to take that? I could take that? I would just say that. Uh, I don't know if everyone here feels the same way, but when I do work, if I feel lazy, I don't use a, I, yeah, it's, it's contrary to what a lot of people think. It's because when I, when I feel lazy or when I, when I do work and I involve a, I, I have now 1000 options to compete with and decide between. So when, when I'm kind of like feeling too tired, I'll just do something of my own and then submit it. Now, I don't think A I allows us to kind of get better work. In fact, the demand is there for us to do this explosion of A I and then either pick really, really well or I do it because you know that if you don't, if you just use that your competitor is going to do the same. So the market forces are going to force you to out innovate A I actually. Mhm um There was another question towards the back. Ok. Yeah, this one here, please. If we could just get a microphone. Thank you. Hi, my name is David from paypal. So I'm just interested in how this gaming industry is evolving. Given, you know, generative A I and I've read news about how certain companies have business models that are quite brutal, right? Really laying off half of the artists, especially those who are strong in two D art, right? You have trained for years only to be let go. And now using the company just uses ja I just keep the best artist there just to monitor the and, and you know, make them come up with graphics or add ons and stuff. So I want to seek your views on how this um you know, gen A I is impacting this workforce, right? And what are the solutions? And do you think it is ethical for, you know, some of the managers and the top horns in these companies to lay off all these 22 D artists? Yeah, great question. So the way I see it is using Inworld, we don't replace people, we use in worlds as a tool to help creatives reach their goal. So for example, if you're a narrative designer and you're creating a character before you had to write every line, the character would say now what you have to do using our, our tool is you set up the character's personality, backstory, emotion, all these parameters, basically, you customize the character and then that character will speak using the personality that you've assigned to that character. And because they can give dynamic responses, now, you don't have to make a linear game, you can make story even even happen and all the characters will react naturally to it. So our tool doesn't uh risk the jobs of narrative designers. Instead it gives them more possibility to be creative rather than change rather than writing the individual lines, they now have to create the entire personality of the characters and then let the characters uh speak naturally. Does that answer your question? Yeah. OK. OK. Um That, OK. We can do one more, very quick question, please. And then uh if we can get the mic very quickly and then a very brief answer. I'm April Sander, who I'm from Sando education. My question is simple is how do you feel about the future with A I to create less uh individualized atomized individuals? Will it actually increase greater communication amongst people and truly reflect greater creativity and culture of people as people, as, as community, as, as as humankind? So how will that impact human interaction essentially? And he says, honey, do you want to take that very quickly? Um maybe a slightly cynical view. But I mean, if you understand that the basis of A I is that it pulls from existing material that's in direct opposition to the pursuit of originality, which is what most creators are here to do. We don't want to copy. I mean, even if the even if we end up, you know what I mean, referencing the intent of originality is the aim. So I guess if you have that in mind as the basis or kind of a guard rail, then yeah, I think it will definitely help with communication in general and just, you know, speeding things up, but you should never replace it. Yeah. Brian Dean Don and honey, thank you so much for joining us. There's so much more to unpack, but we really appreciate you being here.

Presenter: Vishnu SARAN, Co-founder and CEO, Invigilo Technologies

Brainstorm AI Singapore 2024: Saving Lives With AI

I'm Vishnu and I'm the CEO and co-founder of individual A I. So individual in Latin means look after and the whole premise of the company is to look after worker safety. And that's a topic which I'm going to discuss today. How do we use A I to help improve workplace safety? And what does this mean? A lot to me, I grew up in a family where uh my dad is in manufacturing. I have relatives in construction and safety has been a common language that we have spoken from day one and it's been a big problem that's been bugging many countries across the world and it is something that we have to address with or without A I. So this is a typical construction site, a lot of high risk activities. Um busy manufacturing plants, warehouse and oil rigs, one of the most dangerous locations in the world. And what is the one common thing across all of these industries? Every single industry has shocking statistics of accidents. One out of four workers in this space suffers a major accident at least once in a lifetime and this is crippling to the families and the companies about 2600 people are dying every single day because of this, uh you know, the dangerous nature of these industries and this affects personal lives, affects the people, families and a lot of other personal implications to the workers. On top of this, from a financial standpoint, about 2 $50 billion is lost every single year because of workplace fatalities. And why is this happening? You know, we we're living in a progressive world with a lot of safety protocols and things happening that help improve safety standards. So me and my co-founder of in vis we went down to oil rigs, we stayed there, we stayed in construction sites, we stayed in deserts, uh you know, really high risk places to really understand what is happening. And that's how individual was born. And the key problem that these guys face all the time was a severe lack of visibility on what's going on because sites is so big, you can't really know what is going on everywhere. And secondly, you know, when there's an inspection going on, workers behave a lot safer, you know, they automatically become good behavior but the moment workers are not around. So the moment safety manager is not around, it goes back to uh you know, lack of following protocols. And on top of this safety enforcement and patrols are very manual tedious and time consuming. So a bit more of myself, graduated from the National University of Singapore in 2019, spent one year in Silicon Valley doing start ups and interning in a few fast paced companies. I co-founded with my partner in 2021 and about four years into the journey and happy to announce that we have been getting a lot of traction and recognition on regional and local media. So moving on to like, you know how A I has actually been used on the ground by the industry. firstly, you know, uh voice attacks, then you have text analytics and lastly, you have video analytics. All these are different forms of, you know A I technologies that all of us have interacting with, have been interacting with. And even industrial companies have been using even simple things like, you know, Siri or Google or using GP T. These are the most basic and you know, daily use forms of artificial intelligence. But how can we use the same, you know, concept in industrial sites? And that's what I'm going to talk about today um in a bit. So one of the first things that we have done at in vis is to use A I to identify near misses. And this is the concept of using video analytics where we have developed a software that analyzes uh camera streams in real time to detect workplace fatalities and hazards. Things like people walking close to uh open edges, people walking in the line of fire, of moving vehicles, missing barricades, missing P pe and the list goes on. Every industry has their own set of rules and concerns that could be identified with simple, you know A I technology that could be deployed. Secondly, it's been heavily used in predicting uh you know, machinery breakdown faults and for predictive maintenance. A lot of industrial work sites have tons of machines, but there's no way to monitor the health of the machines. So by simply deploying sensors, analyzing the sensors using machine learning algorithms. A side team now has visibility on what you know what to do next. And you know, we have also started using A I to recreate safety lapses, generative A I to create algorithms or create safety scenarios, mimic the various types of hazards, accidents that are not possible in real life. So this is then how we used to teach the A I to detect such things in the future, although such events have not happened before. So this is a way of like creating fake data sets for the algorithms to learn on top of it. So with that, you know, how does A I and what does it mean to workplace safety? It changes more from a reactive space, a more proactive way of detecting safety hazards. It centralizes the experience of numerous safety officers across the world. Imagine one central brain of safety managers uh from thousands of years of experience and you know, various industries across the world. And it's also a safety officer that never sleeps and does not make many mistakes. But at the end of the day, you still need a human to verify and take the action. It doesn't do the whole preventive action for you. It just informs you of what to do next. And with that, you know, I like to say that A I is here to stay and being used in the right manner. Having a very user driven approach, staying with the customer on the ground is how you understand and really tweak A I to help improve workplace safety. And with that, thank you so much.

Vinay AWASTHI, Managing Director, Greater Asia, HP Dahai LI, CEO, ModelBest  ST LIEW, Vice President, Qualcomm Moderator: Jeremy KAHN, FORTUNE

Brainstorm AI Singapore 2024: Keeping Up With AI’s Computing Demands

I think it's going to be interesting panel because all three of our panelists are in some way involved in the attempt to kind of bring A I to the edge as one potential solution for the hardware dilemma that A I poses. I'm actually going to start with, with ST I'm curious what strategies you guys have in place at Qualcomm around this. Um And, and how important it is for you to sort of integrate A I at the edge into to the solutions that that Qualcomm is working on. Well, thank you for the good questions and it's really nice to be here. Um We believe that to harness the full potential of A I and A I has been adopted in all industries in almost all walks of life increasingly so, um it needs to be intelligence everywhere. Um by having intelligence everywhere, we use the terminology of hybrid A I. So you will share the A I workloads whether at the edge or at a cloud and as such, um the full potential of A I can be gathered. And also there's a lot of uh there are all the good benefits that come from it because for instance A I that have been performed at the edge can show a lot of personal security and confidentiality, not to mention uh you know, it free up the traffic going up the cloud all the time. And on top of that, there's a huge economy to this as well because edge devices, edge devices are many, many things, cell phones, smart cars, P CS these days and thousands and millions of IOD devices. And it helps to make sure that that distributed way of processing A I will also tailor for the specific A I where it is needed. So, so we are, we are we are we are committed to that. And that's why one of the things we have done recently is we actually launched our Snapdragon X elite chip set platform for the PC. We believe 2024 is when the PC industry, the PC is being reborn A PC is a pivotal portion of the entire um edge A I and not to mention of course, smartphones and XRVR and so on and so forth. So we believe this is the right thing and the important thing to go. And there is a big need for a lot of the consumers enterprises, small medium size companies as well to upgrade a lot of the devices to be able to do these uh A is at the edge. So I think business opportunities there is also helped to, you know, um involve more people into the uh digitized economy of the future. So that's what we are well suited, you know, Qualcomm is well suited to do that. Excellent. Um Do can you tell us a little bit about model best and and about mini C PM, which is this model you develop and how this kind of represents a step in the direction of these kinds of models that can actually run on on edge devices? OK. Thanks Jeremy. Uh uh I'm very happy to be here to representing uh model best. It is a premier A I start up in Beijing model best is all about creating uh the top level big model and we really value efficiency. Uh We believe that efficient model means better performance, uh smaller uh size and low less cost. But with the same capabilities. Yeah. Um and uh even some folks in the media class Chinese Mistral because we uh because the automated pursuit with high efficiency is the thing we have in only in common. Um uh instead of just, just comparing uh the parameters and the performance, we focus on the knowledge density of the big model. Uh uh That's uh first thing we, we very, very uh very focused on. Yeah. Um uh OK. The star product, our star product is a mini C PM. It is a bunch of uh uh super light uh models that you can run on your phone. Uh um But they pack a punch with bigger ones such as GT 3.5 GT four V or even so future, maybe GT four. these models have been popular all over the world in the past half years. Yeah. Yeah. And you said we were speaking before and that he was saying you, you've released five different versions, updated version C PM just in the last six months is that we, we have, we, we move very fast. Uh We have released 45 versions of mini C PM in the past half years and uh we will uh release uh two versions uh additionally uh in fuel in risk. Oh, wow. So this is technology moving very fast and I think what you can do on devices rapidly catching up with workloads that before you could only run in a in a data center. Vinne, I want to turn to you and, and HP you're known for, for P CS, you're known for printers. Where do you see A I as you know, kind of fitting in with, with that kind of hardware? Well, I think first and foremost, great to be here and thank you for the opportunity. Um If you look at the A I story so far, I mean, I mean, it has mostly been data center and cloud centric right now. We do know that while it's a very powerful story, it is it has limitations in terms of how much it can reach um what we see is the proliferation of A I in a massive way over the next three or four years, every person in this room and everywhere in the um in the world who is using a personal computer right now will find that in order to use A I over the next 34 years, as many of the use cases emerge and perfect themselves, your current PC generations, they will be all obsolete. And for us to really take advantage of A I to every individual, every small and medium business, every large business government. Um you know, this is going to be a massive effort to provide um A IP CS because at the foundation of A I is the need for driving a very order of magnitude higher computation. So, you know, today's P CS, there is a term now we call trillions of operations for second tops, right? Today's P CS are anywhere from, you know, 6810 and we are not talking about 40 tops, 55 tops even, you know, higher than that. So our drive is to be able to provide um a huge portfolio of products P CS, printers, communication equipment that will allow people to be able to use A I models at the edge in whatever use cases they have. Now, it's interesting um when you arrive at Microsoft released this early version of an A I PC, um but it had to sort of record everything you did on the PC and some people were worried about that even though they, they were trying to sell it as more private than having to upload data to the cloud constantly. Um, there was still this concern about security. How are you thinking about this at HP? Well, I think security is going to be extremely important if you look at it, I mean, why is security important? Because as any computational power, any software, it's a tool, you can cut it both ways, right? Um If good guys are not going to use security and A I and the latest tools for the bad guys are going to do. So, so what we are doing is developing models that actually learn from the various attacks that continuously happen and those models will be A I trained. So we have a suite of security products from, you know, bios all the way up to the OS, we call it food security. And that is really based on A I. So it's just one example about how we have to be as responsible technology providers, provide technology that can keep people safe first and then allow them to do all the, you know, enjoy the benefits of the new technology that's going to come. Great. I'm going to ask for questions from the room in a minute. So please think of your questions and raise your hand and I'll get a mic to you. But first I want to, I want to talk here, you talked about how popular mini C PM has been. You also have this interesting case where some researchers at Stanford plagiarized the model or sort of built something that seemed very based on mini C PM. Can you talk a little about what happened in that case? And I'm curious whether this is sort of a risk in general as we move to smaller open source models that, you know, there will be this proliferation, it will be very hard to control what people do with them. OK. It's really interesting. Uh We released a new multimodal version of mini C PM uh that comparable to GP four V on the uh multimodal site uh in on May 20. Uh It's, it's very cool and the feedback from the open source community has been aware. But then on May 29 some people on the github homepage point out to us uh that there was a Stanford project has copied us. Uh And the team claimed to have built a better model with fewer parameters, which was 101% of GD four V, which is true. And the little training training cost only $500 which is not true. And the uh smaller as, as small as smart as the GP T four V which is true. Um After comparing uh we found these models are almost the same, the American one could even understand the Chinese Asian text, which was merely found in the public between data. So as you know, what happened next, uh The Stanford team apologized publicly after uh getting caught out of the plagiarism and they deleted the program uh the project on the github. Um This incident, uh We think it's not uh does not represent uh Stanford officially because it's just uh initiated by a few undergraduate students. Um And we forgive them and uh uh remind everyone to give credit to where it's due in the open source community. And the incident uh just make us uh appreciate the open source community even more. Uh We are thankful for the support to resolve the issue so quickly and fairly. Um And the model best has always been active in the global open source community. Um We have created created uh the open BMB community which has been the biggest open source community uh in China. Uh And we continue to work on projects, open source projects and uh cooperates with other communities to uh advance model technologies, right? I want to get for a question from the audience in a minute, but I'm going to ask a slightly awkward question, which is right now, there's a lot of concern about running these large A I workloads in data centers on the energy consumption. And a number of companies have Microsoft and Google in particular have said, you know, they've been thrown off their, their track to net zero because of the data center growth that they've had. And so some people are looking at on devices, oh, we can avoid all that. But if you, if you look carefully at where some of that additional carbon footprint has come from. It's not actually the energy used to run computations in the data center. It's all the energy that went into the production of the chips that are in the data center. And then there's also concern about, well, where are you running these models? And so if we're pushing everything onto device, my question for Qualcomm is what's happening in terms of the energy intensity of carbon intensity of the manufacture of Qualcomm chips. And then also if we're pushing everything out to consumers and the energy is being used, you know, when they charge their PC or their phone, isn't there an issue there? Because a lot of those consumers might not have access to renewable energy as their source of, of home power or office power? Well, I think first of all, I, I think that we shouldn't say that everything is pushed to the edge. There is a very close collaboration between the cloud and the edge. In fact, if you look at the entire eys I, I would say that uh you know, Qualcomm has always embraced new challenges and new technologies from three G, four G five G. And because of the Connectivity foundation that five G has brought, it has enabled us to provide the options of processing some of these A I at the edge, but you still need the cloud to do certain, you know, it's very efficient to do like massive learning development stuff on the clouds. But then there are minute little things that make sense to do it at the edge. I think the, the, the tricks and the intelligence and the innovations is what to do where. So you, you have things that will happen on the edge, on my little cell phone, my P CS in my car that are not even an option not to do it there because you need that latency, you need that short latency, like smart cars, you need to have that response immediately, right? So there are these things that are just absolutely make sense to do it there. And then as you grow it bigger, the cell get bigger and bigger, you go do the intelligence of the edge and then more edge and then to the clouds. So I think there is that intelligence that with the with the liberalization of the technology and know how and empowering so many smart people to do the right thing, there will be a good balancing point where I think the the industry and the uh the developers and companies will write the will find the right balance to where it is. That's why I think that the model that will really reap the benefits of A I is something called the hybrid A I which is going to be cloud devices and so on and so forth in a, in a network questions from the audience. I have a, have a question, if not, I've got more. So don't worry, I'm going to ask, I'm actually going to turn to Vin A also and ask on the sustainability question. Um It, I mean, you just said everybody is going to going to need a new PC uh in the next few years and maybe we're all going to need new phones too because we're going to need that. That latest elite X shift from Qualcomm. Doesn't that create also a sustainability issue? You're going to have all these people discarding these old devices, you're going to have an E waste issue. How is HP kind of looking at that? Because I know you have a, you have a firm commitment to sustainability, but does this throw you off of that? It does not actually if anything, we are accelerating of a commitment to our sustainability goals. And because you know, our sustainability uh framework is very broad, we, we look at how to make devices more efficient, which means today we are working on devices that can actually have truly 24 hours, 48 hour battery life, you make them more efficient, you know, because there is less loss of power when you are using those devices. That's one part of the framework. The second part of the framework is using sustainable materials as we build our devices. So every PC, every printer today now has recycled material, not just plastic but recycle um material that goes into the chips, recycled material that goes into the casing. We are also making our factories more efficient. So zero waste factories is our goal. Last but not the least. It's a very important part for us to also work as what we call renew supply chain, which means we can take devices back, renew them and put them back into the market, which is because there are a lot more people in this world today who need computational devices than those who actually have computational devices. So for us to work on making our products more sustainable and also our supply chains, more sustainable, making renewed products back into the market. I think that's our framework and we are very committed to that. And does A I have a role to play in doing that making your operations more efficient or find, you know, energy savings throughout your operations? Absolutely. I mean, if you look at today, we are talking about lights out factories, those factories will run because they are enabled through A I. We are talking about devices that will be optimized, whether it is, you know, eliminating the processes that are unnecessary and thereby increasing the battery life. It's all being driven by A. So it has a lot of use. Now, we are the same thing. Well, you know, the, the chip that we just introduced. And because of our heritage in in the very efficient computing and very very efficient power consumption, you are going to be able to do more things in shorter period of time. And imagine that you, you can use these A I models after you really allow the thousands and thousands of developers to think about how to solve some of the problems you just posed, you know, so I think A I with the right tools with the right liberalization and uh you know, open up of the uh of the tools will enable a lot of solutions to solve the problem. Some, some of them you just mentioned, that's fascinating. And while we have Dahi here, I want to ask it's slightly off topic. But one of the amazing things about mini C PM and some of the models you've worked on is their ability to, I guess translate these ancient Chinese calligraphy, these Tsinghua bamboo strips. I don't know if you can talk a little bit about that because that was also one of these tests for the model. And I think it's also one of the ways they caught that, as you said, the Stanford plagiarism because their model turned out could do this too. Which was very strange if it hadn't been using your training. But yeah, the the Tsinghua bamboo split is just uh Chinese Asian uh rep uh that uh uh t economy, uh a t friend uh just uh sent to Tsinghua. Uh So uh there is no uh digital information on the internet. We just take a photograph and use it as the uh fine tune um picture uh to build it into our model. So we very make sure that any other model should have not ability to identify the, the Tsinghua, the character on the Tsinghua. So that's the reason we uh quickly identify the model of the excellent. Well, we're out of time, but I want to thank uh ST and Dahi and Vin for being here. It's great to have you. Thank you so much. Thank you. Thank you.

Teodora DANILOVIC, Prompt Engineer, Autogenai 

Brainstorm AI Singapore 2024: How to write effective AI prompts

So in the short time I have with you all today, I want to demonstrate how smart prompting leads to smart outputs and hopefully give you a few practical tips and techniques that you can take away and use for your own prompt creation process. So I work at auto A where we help organizations around the world write more winning bids, tenders and proposals by leveraging large language models and linguistic engineering. I get a lot of questions around what prompt engineering actually is and what I do in my role. So I thought it would be useful to clarify that. Can I get a quick show of hands who here has used chat GP T or any other similar platform for their prompting? Wonderful everyone. Um Perfect. So what most of you have been doing in those instances is called prompt, crafting. Prompt, crafting is when you interact real time with a model and give it a prompt. For that individual instance, you receive useful and relevant responses, but you wouldn't necessarily expect that prompt to work on any other piece of text that anyone else would use. So, prompt engineering is curating prompts which produce replicable, reliable outputs to fulfill a specific function whilst continuously objectively measuring and improving them. It's about setting up frameworks that scale well in the future, with any unknown input all the time, there are many prompting techniques out there. But these are some of the most popular today. I will demonstrate a few of these as we consider their benefits and their drawbacks. And we'll do this by focusing on one task which is extracting and classifying things from a data set. I'll be using autogenerated platform today to demonstrate these prompts. So the feature that I'll use, it offers multiple options of output for each prompt. So you'll see three options on the right hand side up here, we have a zero shot prompt. It's an instruction with no examples. It's what everyone does the first time they interact with the large language model. It works well most of the time, but it does have a few drawbacks. Sometimes it can lack a nuanced understanding of the task we're trying to achieve. And we'll see that in practice today. So in this example, I bought it to classify a piece of text and that piece of text has a debatable output. So as a human, I would expect the product arrives late, but the quality is excellent to generally be a positive statement. Although it has some positive and some negatives in it, it's weighted heavier on the positive statement. We can see that the model has given neutral for all three outputs. So for this case, zero, prompting doesn't encourage that nuanced understanding and I probably wouldn't trust it for even larger sets of data here. The model lacks understanding of what I mean by each of those sentiments. What does positive mean for me? What does neutral mean for me? In the previous example, I didn't have enough context. So one way of providing it with more context is offering examples of what you want. This is called multi shot prompting a shot being an example, chain of thought, prompting means asking the model to think step by step and show its reasoning. So I've given it pretty much the same prompt, but I've given it three examples of what I think a positive, negative and neutral statement is you can see as we transform the text, the outputs are far more new. The first one says positive. The second one after explaining its chain of thought where I've instructed it to think step by step also concludes to be positive. In the end, it's far closer to what we're looking for. One thing I will say is that you should be wary of bias with multi shop prompting. In this example, the model might take it to mean that positive statement must always talk about product quality or it always must mention that the site was confusing or that it always has to have one positive and one negative side to it. If you're using multi shot prompting for larger sets of data, you have to make sure your examples cover all bases. This can be quite difficult to do as you have to think of every way that something can be interpreted chain of thought also helps in this scenario. Because if I can see the models thought process, I can see where it went wrong and it can help with my model debugging multi short hair was really great for improving this task and giving the model some more understanding of my intention. But sometimes I want to do something a bit more complicated and I don't want it to give me just a one word classification for something. Let's look at how we can introduce even more complexity with some prompting techniques, prompt training or multi step prompting is best for complex reasoning tasks that cannot be instructed in one go. It ensures you're working on the best piece of text at each stage and it doesn't leave room for model inconsistency. It makes sure that the potentially conflicting instructions don't interfere with one another. Let's say I want a more complex analysis of some sentiment on a larger body of text. As a human, I might break this process down into a few different steps. First classifying the statements as a whole, then extracting themes from the statement and then grouping those themes. This type of breakdown also works great with prompting. So for the sake of brevity, I'm just showing the output of the first prompt that I gave the model. The prompt instructed it to classify a list of customer feedback into sentiments using the same multi shot prompt that I showed you guys before. Except on a larger set of data. There's around 2526 pieces of customer feedback. Here, this is the first prompt in this chain. The second prompt in the chain is a really good zero shot prompt up at the top here, you can see we've asked it to identify all the themes in the following list of customer feedback. Your response should be an exhaustive list of accurate and relevant themes with a number beside them indicating how many times it appeared, do not write the same theme out twice, feedback prolong. So I think it's important to note some key parts of that prompt. So I've asked it to identify all the themes and then I repeated myself and said it should be an exhaustive list. Repetition is always good. I've said that it's a list of customer feedback which provides the model with context of what I'm talking about. I've asked it to be accurate and relevant. So it's clear that I want the themes only to do with customer feedback. And then I've said how I want the structure to look, you can see with all of the options. The structure is exactly how I want it to look. The answers are exactly what I'm looking for. So I'm happy with this and we bring it over into the editor. So as we move on to the third prompt in the chain, you can see how with each step, the output is a combination of everything that came before it context which is implicitly woven into the answer gets carried on to the next prompt. The prompt that I've given it here. The third prompt in the chain is classify the following themes into positive, negative, neutral or other categories. Your response should have the sentiment as a heading followed by the themes which fall under that sentiment under each theme, write a brief justification of why it falls under that sentiment. And then I've given the list of themes, as you can see on the right hand side, this is far more nuanced, accurate and useful than the first thing we looked at. This is this output is actually a combination of all the techniques we've looked at so far and we could not have gotten this type of information, which is this thorough and this accurate through only using one of the techniques. Once you've got an output that you're happy with, the possibilities really are endless. You can translate it into Jason, you can play around with a toe, you can turn it into a powerpoint presentation and much much more so to conclude this short segment, simplicity is always best. Although a single shot prompt wasn't nuanced or accurate enough. For example, today, it is most often the best choice. The last three examples I showed you of turning it into Jason or translating the tone. All of those used a really good successful zero shot prompt. A prompt should be direct, unambiguous and relevant. And each of the techniques we've gone through today are just different ways of ensuring that the prompt meets those requirements. I really hope this offers some insight into what prompt engineering actually is and gave you some tools that you can take away and implement into your prompt. Crafting. I think we might have time for one quick question. Is there any questions in the audience? Yeah, we have the one here. How do you think about it? And do you use any tools to refine your prompts and how? Yeah. Yeah. So that's, that's actually a really interesting question. Um You can prompt models with how you want them to refine your prompts interestingly. So you have to have the techniques in order to instruct the models and how you want your prompts to look. Anyway, sometimes if I'm lacking inspiration on how to begin writing a prompt, I will definitely use a model. I'll be direct, unambiguous relevant, clear about what my parameters are clear about what my instructions are. And often it gives me a really good framework. The only problem with models are, it's sometimes a bit more nuanced than that. And you're looking to write a prompt for a specific use case. So I know my target audience. I know my, my customers and I know what they're looking for and I can try put that into prompting. But ultimately, it's your subjective opinion on whether you think it meets these metrics that you set out. So, yes, absolutely. You can use other models, not only Claude but any of the providers to give you a first draft of a good prompt and to better your own prompt. Awesome. I think that's all the time we have today. But if you have any more questions, I'll be around afterwards and please feel free to add me on linkedin and ask me any questions there. Thank you so much.

Seth DOBRIN, Founder and CEO, Qantm AI Tim ROSENFIELD, Co-founder and co-CEO, Sustainable Metal Cloud Moderator: Clay CHANDLER, FORTUNE

Brainstorm AI Singapore 2024: It Ain't Easy Being Green

Brainstorm AI Singapore 2024: Will AI Help Combat The Carbon Crisis— Or Make It Worse?

So, um I want to just kind of follow with a couple of questions for you, Tim, because these were really uh impressive. So you can, you can basically get carbon emissions, electric power consumption uh and, and spatial requirements down by about 50% for each of those three things. And I couldn't quite hear you back. But did you mention the recent kind of metric that, that uh kind of benchmark this? Yes, I didn't actually mention that on the presentation, but indeed, no, it, it's one thing to come out with a system that um you know, is very energy efficient. Obviously, we live in a world of um of intense commentary around sustainability. And there's obviously the propensity for greenwashing. It's just a fact of life. So it was very important for us that we could demonstrate to our users and frankly to folks that we're in dialogue with who might use our technology, that this system despite being generationally further ahead in terms of its consumption of energy was as performant as it could be. So to your point, Clay, um we worked with a industry association called ML Commons, ML Commons um For those that don't know, is um I'd kind of describe it like the Olympics of A I uh for, for geeks. Basically, it's an open source uh forum. It was founded by NVIDIA and Google and Microsoft and there's about 100 and 25 members now. And every six months there is a competition to run different A I models and see how fast your system runs them and forever and a day that's been the main metric as to how this Olympics are judged moving forward. Um And this submission that we just did ML Commons said, look A I has a sustainability problem. There's a green problem with A I. So why don't we start tracking power? How much power does it take to run these models? The end output is we've been the first in the world to submit, not just the performance, but the power benchmarks and have that verified by the industry. And it's, it's a very exciting moment for us. It's great. You know, um when we were visiting this data center, you made some comments that I thought were really interesting and I thought maybe I might ask you to share with the group here. Um That the reason you came to Singapore was kind of the of the uh you know, the, the, the paradox of plenty. You know, we often feel, you know, hear the story about how uh countries that are blessed with lots of natural resources actually don't develop that much because they don't have to. Singapore is the opposite of that and maybe you could say sure why, why that influenced your decision to come here. As I mentioned, I'm, I'm, I'm from Australia um originally and Australia is a big country with a lot of energy. So that's sustainability and, and uh efficiency has never been high on the agenda. Singapore is a, is a, is a perfect environment to put, to put, to work. The thesis that we need to do better. Singapore has power constraints, it has land constraints, it has a very strong drive at government policy level and within in within industry to do better on the sustainability front. So for us moving our business to Singapore and setting up Singapore as our global hub um was a good test because we've been able to deliver into Singapore more A I capacity than um has ever been available before in a private cloud like ours. Thanks to this ability to use half the energy. So that's one of the ways that we've actually been able to showcase in a constrained environment like Singapore, this type of technology is, is needed. So it's, it's interesting to me because it's, you know, this observation that it's in precisely an energy and, and land and resource constrained environments that you, you have to radically innovate. You can't just tweak the traditional air cooling system and get, you know, three or four or 5% increases. Uh a year, you've got to do something that's going to radically cut, uh, resource uh, requirements. Uh, so it's kind of like, you know, if you can make it here, you can make it anywhere. Right. Exactly. And, and, and, you know, I, I absolutely love Singapore. I live here. It's a, it is very much that, that ethos, if you can, if, if you can prove it in Singapore, it's hot, it's constrained. They want to do better on sustainability. Um You can prove it to the world. Seth, let me get you uh involved here. Seth. For those of you who don't know is the former uh A I officer at IBM. He's written a wonderful book which is available, I think in the lobby if you get a chance to pick one up. Um And uh is really one of the most uh kind of far reaching A I thinkers in the world. We're delighted to have him here. He has also, he is also the founder of a start up that is investing in sustainable A I technology Seth. Maybe you could say a little bit about that and why you have chosen that path. Yeah. So in addition to quantum A I, which is a consulting company, we also have a venture fund called One Infinity, which is part of the reason I I came here is to, to raise for that and we invest in responsible safe and, and Green A I. Um And investing in green A I is important. Uh A because there's a lot of greenwashing going on. Uh You can get on, on quote unquote, green clouds that's moving on green CRE carbon credits, which is multiple counting of it's not really green. Um And so just like a lot of things aren't really A I. Um And so we really thought it was important. My partners and I to look at how can we support true investments in things that are green, like really green new hardware, uh new paradigms and how you support A I uh and, and workloads around A I. So some of them like, like with Tim's building uh completely new ways of thinking about how you operate on silicon. So uh moving away from traditional semiconductor technologies to new silicon technologies that are 1000 times faster than today's GP US or new mass compression algorithms that are 50 times faster uh than traditional uh matrix multiplication, which is how uh language models run today or even things as simple as I'm going to have GP U as a service. And some percentage of those GP US are going to be available as certified green clouds where they run on wind or solar or oh my gosh, if you put it in Iceland, it can all be geothermal and not only is it completely carbon neutral in the wintertime, it's carbon negative because they use it to heat the housing and the environments around them. And so there's lots of very simple things that start ups are doing that are really truly green. Uh And so that's why we decided to, to start investing in, in platforms like this and, and by the way, you can make a lot of money on them. OK? One of the things that Tim and I discussed was I hadn't kind of fully realized is that part of the issue here is the people that operate data centers. Um they're not engineers for the most part, right? Their, their uh the real estate investment trusts uh and private equity funds. That kind of think of these structures as office buildings or more or less an Amazon warehouse or something like that. But as Tim was pointing out, you really have to think about the engineering and think of them as factories in the way that an engineer uh uh would um how, how much I mean, if we can switch to these kind of liquid immersion cooling type technologies and away from the traditional air conditioned uh uh model, what do you think Seth about that? Is that, is that going to be a solution that will really move the needle in terms of power and space consumption? You know, I think there's multiple ways that you can, you can tackle this, right. One is, you know, there's a lot of, you know, traditional data centers are not just air cooled, they're water cooled, water pooling creates problems of itself. Right. If you go and use chat GP T today or an, or claw, you know, I don't know how many people, you know, this, but for every 25 to 50 prompts, depending on how the, how big they are. You use about a half a liter of water just through evaporation. Right? I mean, so, you know, they're liquid cooled more efficient than a and then air, but you're still harming the environment. Um And so we absolutely need new paradigms of cooling. You know, people say we're gonna, you know, consume 70% of the world's electricity. You know, we've been running out of oil for 50 years. We're not going to run out of electricity. People like Tim are gonna invent new technologies where uh and we're gonna hopefully fund some of them uh that are gonna get us to where we're gonna continually find new ways to make sure that that doesn't happen. So it's gonna be things to like new cooling. It's gonna be new hardware paradigms, it's gonna be new math paradigms, it's gonna be moving away and people talked about this earlier today. You know, these massive large models are not what the technology was built for. They're cool, they're fun, but that's not the solution. That's also gonna help the power, it's really small models, it's more efficient technologies. And so we're not gonna have this problem of these massively large models used to all these powers because that's not what the solution. Yeah, that's very interesting. So, a couple of things I hear there, um, one is, um, well, so on this, on this uh problem of, of kind of water consumption, you, you sort of figure out how to recycle the water to make that more efficient. Well, you don't use water, right? You use, we, we use wastewater so we use wastewater and, and, and I'll describe a little bit, you know, for, for a little GG for a second here on how it comes together. But this con of that, that you talk about clay, which is that, you know, traditionally the data center, the end to end energy into knowledge, whether it's cloud, whether you're using, you know, whether you're doing a Google search or whether you're using, you know, whatever Apple Maps, there's all these things in between, which is this, this point, you know, and, and you just said it too. There's this point behind it, the actual factory that most folks don't see. We're, we're using phones now and, and today and you just don't, you don't feel it and you don't touch it but its real world impact. So this notion that there's actually been a silo, there's a data center who builds a building, there's a cooling engineer who builds a cooling system which involves chilled water. There's a computer manufacturer who builds a computer, there's a chip designer who designs a chip and each one of those people says to the other one or you go figure out your part about this, you go figure that out. You, someone will figure that out. You know, when Jensen comes up with a chip from NVIDIA, someone else figures out the cooling and that's, and it's been fine up until recently. The problem is that graph that I put on on screen before. Um There's this insane amount of energy consumption now going into these chips and a vertical system like ours is the way that we get over that to your point. Clay and question about water. Yeah, most data centers, whether they're using it for direct chip cooling or whether they're using it for air, they use chillers and that a real energy hog, you know, especially in this part of the world where it's hot and humid. So when you use a chiller, you're really over cooking the amount of energy you need our system, our system takes the building's effective waste water. So it's called condenser water. So after the, after the water has done its job in the rest of the data center, it goes out into an Evaporative cooling tower and it gets recycled. We take it at its hottest point before it goes into that cooling tower, we bring it in to our hypercube which is a self contained system and then it does its thing. So in that sense, technically, we use water, but we use water that was already in the loop and we don't use chillers. So that's the, that's the energy saving. What this kind of suggests is that um there's been driving the chips, you know, Moore's law, which we talk about all the time is kind of well at work in making them faster and better and smaller. But that there hasn't been up until now, a real Moore's Law type dynamic in the way that we think about data centers, we just use the, we build the big ones out, keep on building them now. And that's what we need is we need to start applying engineering and science to, you know, condensing those I I think to a certain extent. So data centers got smaller for a long time, for a long time, data centers got smaller and smaller, so much to the effect that, you know, there are a lot of places where they were trying to figure out what to do with the space in data centers. Um But now that A I is really taken off data centers, there's in some data centers are still a lot of room, but they're consuming more power because the chips are more power hungry. So in a lot of places, it's not a space issue, it's a power and heat issue. And so we shouldn't talk about a space. It's really a power and cooling or slashing issue. And it's not that we can just put data center, all the data centers in Iceland or or Sweden or, or, you know, Denmark? Well, I, I would challenge that actually because do we really care about latency when we're having a conversation? Every, and, and we need to really think about, when do we care about latency when we're training models? We don't care about latency when we're using models. When do we care about latency if I'm having a conversation? So, if my employees are having a conversation with the system, do I care if it takes a half a second versus 1/10 of a second? So I really care is that worth impacting my company's carbon emissions? Right. Right. But that, that's a really interesting insight because it, it uh it's a focus, it shifts the focus away from the, the operator of the data center or the manufacturer of the chips, you know, to the actual user of the data. And what you're suggesting if I hear you correctly is that we all need to figure out a way to be more thoughtful about the trade offs involved in, in, you know, using it's part of the decision, right? And getting back to what you said about my book, right? That's part of my conversation. Look, it's part of the architectural decision, the conversation you need to have when you're designing each use case and it's part of the human conversation, right? When you talk about Human Centered Design, right? Panos was talking about the human, you know, human A I, you got to think about the human, who's the humans going to interact with it? How is the human gonna, you know, how is the human going to interact with it? What is the human decision? How important is that human decision? All of these things play into the architecture which has implications on we're talking about human centered. When we're talking about green, it has implications to how you make that, how green that, that, you know, how you, how would you, you would greeny that, that, that, that, that A I application? Yeah. So it's looking at the kind of demand side and the, and the psychology of the user, I mean, I use, you know, chat GP T all the time and I've gotten really lazy about it. I mean, you know, it's so good, it's so efficient that if I can't think of the right word, I'll use it as a thesaurus when I've got an actual thesaurus that's probably on the bookcase behind me. And I don't, there's nothing about the way that, that, you know, is delivered to me, that service is delivered to me that makes me stop and think actually, you know what, for the amount of resources I'm gonna, I'm gonna burn to just do this. I, I should actually pick out a real book or use Siri on your phone. OK. 11 interesting unfortunate headwind to that though is the um the uh growing um narrative around data gravity and data sovereignty, which is, which is, which because I'm of the same opinion. In fact, you know, a system should be designed for its best use case. You know what we've designed. Our A I factories are not necessarily what you'd use for storage or what you use for general purpose. CP US, there's different, different horses for different courses, but there is a very real and present. I mean, we see it with the Chips Act but you know, it's a lot more out there. There's very restrictive laws heading towards parliament in India. There's a whole lot of countries that are very considerate about making sure that d doesn't leave the country. Yeah, and it might not always be the choice of the end user unfortunately. So there's a policy piece of that as well. Well, it sort of came up in our conversation with Minister C about data sovereignty and if you know, if you are Singapore, for example, you can make the argument. Well, why don't you just move your data across the border to Johor? It's cheaper, lots of space. Um But there's lots of data that you really like. You Singapore want to keep on shore, you know, your national health insurance data, you the defense data, the military data, uh anything that's proprietary commercial data, the big banks, you know, these kind of companies, there are a lot of multinationals who want set up a, you know, a data center here but not have it across the border because they're not sure that it's as stable as, as Singapore is, which has a kind of better international brand. Well, if you're multinational, you can't control data sovereignty laws, there's data sovereignty laws in the Middle East, there's data sovereignty laws in Europe, there's data sovereignty laws in South America and in Africa. And so yes, Singapore can control it, Singapore, but you can't control it for the rest of the world. Um And, and so even outside of individual countries, they're gonna exist. Um And, and some of them are for good reasons. Some of them are just because so popular. Yeah, but, but there are technologies now that you can actually, you know, Federated learning actually works. Now, uh you know, we talked about it for a long time. It didn't really work now. Now it actually works. So if we were to think of just kind of one or two things that we could really focus on and prioritize to fix this problem of energy consumption and carbon emission of data centers, what in your view would that, would that be? I think, you know, when you're, especially when you're looking at generative A, I get back to the root of what the foundation models, which is the technology underneath generative A I was actually built for, look at, look at using small task specific models, so focused models uh those use a lot of energy because less blender energy because they're smaller uh focus on the use case, what problem you're trying to solve and then determine what type of latency you can, you can, you can tolerate. Um And then I think, you know, focus on, do you really need generative A I? What's the right tool for the job? Uh And I think those are the things that you can do today uh Before we develop new technologies. Really interesting. Ben, a very sophisticated. I, I feel like I've learned so much from this uh from this panel. Uh Tim, I know you all are looking at bringing some of your manufacturing capa capabilities to Singapore uh fairly soon. Uh And, and we'll be rolling out, as you mentioned to Thailand, India. And uh what was the other one in Australia in Australia field? Great. So it looks like it's a, it's a promising market for that kind of technology. Um We are out of time. I think we'll have to leave it there, but I want to thank our two panelists for a fascinating conversation about what really is a, is a crucial topic uh for, for artificial intelligence. Gentlemen. Thanks very much. Thanks. Bye.

Presenter: Tim ROSENFIELD, Founder and CEO, Sustainable Metal Cloud

Brainstorm AI Singapore 2024: Cutting Data Center Energy Consumption By 50%

I've got a short amount of time, about five minutes uh here to describe um a little bit about what we do and um how we fit into the A I and energy ecosystem before we have a little chat on stage. Um So I'm Tim co-founder, Co CEO of Sustainable Metal Cloud. I was trying to think of a way to describe what we do in a very short amount of time and I guess I'd, I'd characterize it as we seek to solve A I's energy problem. Um A I has AAA substantial energy challenge and problem um which will unpack over the course of the next few slides. We started life in Australia as you can tell from my accent, I'm Australian um about seven years ago. Uh our mission and our vision myself and my co-founders was as outsiders to the data center and A I or as it was known then HPC industry. Um How could we apply engineering to turn energy into knowledge more efficiently? Now, what I mean by that is for most folks, the data center is a place that data is stored. It's now becoming a place that data is processed these A I factories are actually large engineering challenges. And our, our process was, could we start a refresh? Could we think of what is the best way to do this and to accomplish this? So we started in Australia where we built a 20 megawatt um engineering center. The problem that we're trying to solve for how to turn energy into knowledge as cost effectively and as efficiently as possible starts with heat, heat and cooling. And that is one of the drivers of A I's energy challenge, roll it back one step. And the graph I've got on the screen here, you can see a very clear trend. And what this is showing is that over the last decade or two, we've had real substantial improvements in chip density and chip efficiency. But as we've come to now, two nanometer three nanometer process nodes, we're getting less and less efficient with every step. And this has coincided with the rise of A I which unfortunately uses the most energy intensive and hottest chips out there. So behind me, on, on the graph here, you can see in, in green, we have the power limits of nvidia's GP U series. So this is ramping up at a very quick rate and this is driving a massive challenge in the data center. So more heat in the data center means you need more cooling to remove that heat. And that is why our solution it's quite different. So we use liquid. Um We've developed a liquid platform that you actually take the servers, you put them in. It's called single phase immersion. A full server in oil and use the properties of oil which has a heat coefficient 1000 times greater than that of air to remove the heat from the servers. Our system was designed to be modular to go into any data center and to host commonly found workloads in the A I ecosystem like generative A I training using H 100 omniverse using L 40 s and other types of platforms in a nutshell. What it is is we want to solve for taking any DC like the uh like the strip diagram here. And coming up with a platform that would fit into that data center, whether that was built for liquid or not. So whether it was designed 15 years ago or whether it was designed this year, either way we wanted something that was agnostic to the data center that it goes into. And then on top of that, we wanted to improve the efficiency of the servers themselves. So by integrating the data center with our new type of A I factory with an NVIDIA GP U based server, we could achieve breakthrough efficiency which we'll get into in more detail in the next session. Long story short, by vertically integrating the solution, we cut around 45 to 50% of the energy to run NVIDIA GP U based workloads today, over standard airbase configurations. We've designed this to be containerized so that we can deploy it in an existing data center or we can build Greenfield data centers. What we're doing here in Singapore is we've deployed seven of these systems. We're deploying into India. We, we've deployed into, into Australia and we're deploying into Thailand and we're now looking to other regions and around the world. And the final slide before I leave you is how does this get used? So we're actually trying to take this to um as many places as possible. And one of the ways we do that is installing the infrastructure in the data center, making it an A I factory, but then making it easy to use by allowing others to consume this service as a managed service, whether it's a cooling as a service solution or whether it's a GP U as a, as a service solution with partners like NVIDIA. Um and customers around the world are able to access this today.

Juhi MCCLELLAND, General Manager and Managing Partner APAC, IBM Consulting Justin NGUYEN, General Partner, Monk’s Hill Ventures Chan Yip PANG, Executive Director, Vertex Ventures  Ren Yeong SNG, Managing Director, Artificial Intelligence Strategy & Solutions, Temasek Moderator: Ellie AUSTIN, FORTUNE 

Brainstorm AI Singapore 2024: Opportunities In The APAC Market

Juhi MCCLELLAND, IBM Consulting, Justin NGUYEN, Monk’s Hill Ventures, Chan Yip PANG, Vertex Ventures, Ren Yeong SNG, Temasek Moderator: Ellie AUSTIN, FORTUNE

Jacqueline POH, Managing Director, Singapore Economic Development Board Su Shan TAN, Managing Director and Group Head, Institutional Banking, DBS Group Moderator: Ayesha KHANNA, Co-founder and CEO, Addo AI; Co-chair, Fortune Brainstorm AI Singapore

Brainstorm AI Singapore 2024: Public - Private Partnerships In Ai Governance

Jacqueline POH, Managing Director, Singapore Economic Development Board Su Shan TAN, Managing Director and Group Head, Institutional Banking, DBS Group, Ayesha KHANNA, Co-founder and CEO, Addo AI

Gullnaz BAIG, Executive Director, Angsana Council Evi FUELLE, Director, Global Policy, Credo AI Zee Kin YEONG, CEO, Singapore Academy of Law Moderator: Clay CHANDLER,  FORTUNE

Brainstorm AI Singapore 2024: In Search of A Governance Model that works

So this is a, a very rich and complicated topic just like the region itself. Um I want to start maybe if we could just thinking about the opportunities we heard earlier in the discussion about whether uh uh customized largest language models could be devised for Southeast Asia, that there were all kinds of commercial opportunities that would uh fall in line with that. And I'm wondering if the same logic applies to uh governance models. Um What if you have thoughts as to what the, what the advantages might be of being able to come up with some kind of common framework or maybe we can think about what that framework might be uh down the, down the road a bit. But, but just talk about why the region might have an incentive, whether it's commercial or otherwise to kind of band together. Do you want to take the first shot of that? Yeah, sure. Firstly, thank you for having me on this panel. I sit at a unique intersection being the director for and final Council. We sit pretty close to start out various tech companies being from the tech sector ourselves. But at the same time we engage quite regularly with various governments in the region. And I think what we see is in terms of ASEAN integrity. And there being a digital economy per se by 2030 we could be the third largest economy in the world. If we were one country, there is extensive business potential. But the challenges lie in the fact that it's actually very hard to go across border. So the ability for us to leverage on the true potential that is offered by Southeast Asia sometimes hampered by land of say, for example, simple things like business recognition across border, some issues around cross border data flow, cyber security standards. Um Just to introduce a few that I'm quite sure my co panelists will dig into. Um So if we can get a framework together, there is this huge market out there for us and you do see some bigger markets in the region leaning towards protecting their own domestic market by putting up regulatory barriers. So that sometimes gets in the way of developing a regional framework. Really interesting. So and it's, it's an interesting, it's a, a big opportunity is a market, but it's also a big opportunity for from an investment standpoint. Like if you're a VC investor, the ability to be able to invest in a company, a company that operates across an entire region uh versus just in a single market uh is, is huge. And I I like your point about um you know, it, being the third largest economy, it's not only that, it's, it's large. If you think about it in an aggregate, it's also that it's got great demographics, most of the, uh, in individual economies in it and nearly all of them are right now growing at very respectable high growth rates compared to many of the other world's economies. So, if you think about that down the, down the road, it really, it, it kind of adds up you. What, what's your take on it? Yeah, if I could add to the comments you've already made, I think uh at Creta A I, we're working with enterprises that are highly regulated and they are global in scope and they're operating in these markets all over the world. So not only in this region but globally, they're looking for that commonality. How can they create a governance framework that is gonna stand the test of time? That is not a brittle uh market that only works in one jurisdiction or only satisfies particular risk uh requirements. So we're constantly looking for those elements of commonality. And I think that the benefit of that for businesses is building trust over time in different markets with different uh regulators. Uh but also having certainty with which they can operate and continue to operate in the long term. So I think we've seen a lot of commonality um even between things like the EU A I Act and the White House executive order on A I in terms of what the requirements are from enterprises, things like having comprehensive A I governance risk management systems, transparency reporting, some form of public notice impact assessments and A I registry. Things like this are, are sort of the common elements we're already seeing emerge. How about how about you? I know you work carefully with it. Yeah, so I personally have been working in this space of convergence for many years around data policies and A I and what what and the Academy of Law actually has a think tank, the Asian Business Law Institute that does work firmly in this space. I think it is utterly important for us to really take a step back and look at Southeast Asia as a whole, right? We need to understand the maturity of the economies and also the technology, right? And to I think majority of Southeast Asian nations are likely to be deployed and users of A I systems. Now, the good thing is that there is a fairly consistent set of principles around responsible A I, you can point to, for example, the OECD A I principles, right? As a set of principles. Now, how do you take a set of principles like this, bring it into Southeast Asia and achieve two goals? One to make it a lot easier for companies that are trying to bring products and services into Southeast Asia to very quickly localize the same products and make it a lot easier. The second is of course, how to assist companies in Southeast Asia to also adapt to that and grow out and expand beyond Southeast Asia. Uh We, we have actually a very, very useful platform ASEAN, right. And through ASEAN, it is possible to develop uh a, a governance framework for responsible A I modeled after and looking very much like the OECD princip and use that as a mapping standard so that as individual companies develop their own national frameworks, they're able to map into something like that. The ASEAN framework can be moderated, right, to suit the situation for ASEAN, which is largely deployments, right of, of A I system and also to, to adjust it such that it's a lot easier, the companies and the the the governments working in this space really interesting. I think there are several things there that uh have caught my ear. Um One is, you know ASEAN as a, as a framework for this, there have been, there's an ongoing discussion really about how ASEAN might uh kind of work together to come up with a common framework. Um How, how far along do you think that is? And is it uh is it, is it making progress? Where do, where do we stand right now? I mean, obviously, uh people think about the eu and they think of it as being fractious and always arguing and disagreeing and they haven't seen anything until they've seen Southeast Asia right, much more complex, diverse region and harder to, to come together. So where, where do we stand? I mean, I don't know how many of you here are technologist or from the sector, you know how fast technology develops and then you've worked and interacted with governments. I was from the public sector, um government, public sector don't move that quickly. And with A I particularly the break net speed at which different companies are racing to get ahead and get towards AGI governments are never gonna keep up, let alone a multilateral institution like Asian. So you need a framework but you also very quickly need to develop meaningful public, private partnerships. Um And that can help forward the goal. Um So, so that I'm going to pause back because I can talk and toy about this. No, that's really critical and it's interesting, I mean, Singapore story definitely uh underscores that, but either of you have kind of thoughts about how far along we are in terms of discussions about a framework. Well, so, so iiii I, I'm a glass half full kind of person, right? Yeah. So because I I look at it from the perspective is that from the perspective that number one ASEAN is not like the eu the way we are put together is very different. So we can't expect the cohesiveness, cohesiveness of eu in terms of policy making. Second, imagine if we didn't have ASEAN, right? So while there's a lot that can be said for moving faster for this kind of organization of Association of nations. This is the speed that everyone's comfortable with. It's better than not having something. But I would say um for A I governance is probably is likely to take the same kind of pathway that ASEAN took for data protection. Just imagine back in 2016, the ASEAN Ministers for Tele Telecommunications adopted a set of data protection principles for ASEAN progressively over the years, they built out the data management framework model contractual clauses to assist in the cross border transfer of personal data. It takes time, but we've concerted efforts, uh things can be developed. Now, these are still voluntary frameworks. They are not binding on the government, but at least they exist and once they exist and it's put out by an organization like ASEAN, there is a bit of impetus to say let's make use of them and let's try to be coherent. I can imagine that for A I, the same path might be taken as well. I think the first step has already been taken. The uh the the responsible A I framework has been, has been set out, right? So I think we can expect more of this down the road, but a lot of work still needs to be done by us in the room, right? As business owners to, to really reinforce that message as we are talking to policymakers and governments to say that look ASEAN put up something, right? Can we just make sure that when you introduce national policies be as coherent as possible? Right? Use it as a mapping standard because it helps businesses make sure that their internal compliance program is easy to manage. It makes it a lot easier to deploy products and services across Southeast Asia. So I don't claim to be an expert on, on kind of the ASEAN discussions about this, But I was struck to notice that there was a kind of big delegation of folks from the Eu who trooped to Southeast Asia last year and tried to persuade everyone to, you know, let's link up, let's have a common standard for, for both our markets, not just one market and to adopt our kind of uh GDPR uh framework and then it'll be easier for companies and it makes a lot of sense, but the response was not very enthusiastic from Southeast Asia. And surprisingly, what the Southeast Asian said is actually, we're very suspicious of over regulation because we're afraid that we're going to miss out on some of some of the benefits of, of innovation and these technologies will pass us by. Um So should we be discouraged by that or in a way, can we be encouraged by it? If we think that at least in that regard, all these countries seem to have sort of a common outlook, there seemed to be something resembling a consensus on that, do you have any thoughts about that? Um, I'll jump in but I would love to hear from my colleague here too. Um, we, we're sort of a mix a hybrid region. Right. We're very diverse. Um, we, we love each other yet. At the same time, we're always looking over our shoulder and wondering what our neighbors are doing. Um, so our model has always tended to be a hybrid of both regulate when we need to regulate or we want to regulate and regulation tend to come fast and quick here. We always need to keep a sense of the pulse. Um But at the same time, we like to keep an ecosystem, eh that's open to innovation, to test things out. And we like that because what you see happening in the eu is regulation comes up pretty quickly and to a certain extent it stifles innovation. So how do you do this balance? And I like what's happening in Thailand, I like what's happening in Singapore where governments are reaching out and thinking about how do I partner up with the private sector to think about meaningful applications of A I and I agree with my panels here in Southeast Asia, we're at the stage of application, not necessarily building up the technology from ground up per se. So governments really need to think about what makes sense for my country and to ensure that the infrastructure available in the country supports the application. So, for example, in Thailand, Thailand 4.0 for example, looks at industry application of A I and you know how they're gunning for a lot of investments in Thailand and making use of A I to automate certain technology, certain part of their manufacturing is putting them ahead of the game. Yeah, Thailand is kind of streets ahead of almost everyone in the region in terms of adopting solar. Yeah, but also in terms of data centers, for example, like Malaysia is gunning for tier three data centers to make sure that cyber security integrity, electricity availability supports the ecosystem. And similarly, in Singapore, we lifted our moratorium in late 2023 but plays very high standards for the next generation of data centers that's opening up. So really government's role here is to partner up with private sector and to ensure that the ecosystem supports innovation and development of enterprise. So I want to dig a little more deeply into that because I, I think there's there's a lot to unpack there. But uh do you have a thought about it? I I think just building on what you were talking about uh clay with the campaign for from the eu to sort of um do something similar around A I here in the region as they did prior to GDPR, I think that uh where enterprises are, are trying to do something different is looking beyond concepts like being a good steward of the data and thinking about how A I systems interact with each other and the unique risks related to A I. And for that reason, there's a little bit different approach taken to the regulatory aspect. So the EU has the ability to be united in its enforcement of the EU A I act, which I think is a big differentiating factor for why they're, you've got to have a much more kind of consensus approach here. Just, just one little agenda for that trade mission you mentioned, right? So while they came and as you mentioned, they wanted to no evangelized GDPR, there wasn't any pick up, right? But what came out of that, what came out of that was actually a joint effort between the European Commission, right, and ASEAN. And they came together and they did a mapping of the standard contractual clauses and model contractual clauses that came out of ASEAN standard contractual clauses that came out of eu because they recognize that trade needs to carry on between these two regions. So let's have some effort to try and match and make it a lot easier for businesses. So while there's a recognition that maybe some of these policies are not entirely suitable, but there is a recognition that trade needs to flow, data needs to flow. And is that kind of spirit I think that we need to encourage and I want to leave some time to take some questions from the floor. If anybody's got burning queries, put your hand up and wave at us and we'll, we'll recognize you. Um You mentioned the sort of things that uh many of these, these Southeast Asian nations have in common in terms of what they're looking for is a regulatory framework. But of course, there are individual, you know, specific issues that relate to A I that often kind of put them at odds against each other. Uh Data centers is one, you know, as soon as the moratorium happened in Singapore, uh all the operators rushed across to Johor instead of, you know, in Indonesia started putting up data centers there until there were more data centers across the border than they were here. And then the government went. Hm, maybe this is not what we want. So that's a kind of a friend of his example. The other one that's fascinating to me and I don't know, maybe none of you want to comment on this, but I spend a little bit of time looking in the Philippines and the Philippines is very ambivalent about, about A I because of the job uh uh implications, right? You've got a very big uh population of people who are employed, working at these call centers. And that's probably going to be one of the first places where companies are going to figure out that they can roll out uh you know, interactive chat bots with generative A I. And so they're much less enthusiastic about how, how do you see that is that, are there similar tensions in other, other economies as well? And are those insurmountable or there's a, there's a really good research piece done by Trisha Linkedin where, where they kind of did a study on the extent to which A I was gonna disrupt the employment of an individual or this place or augment or it's completely non applicable? Right. And this study was done on Indonesia. So for the bulk of the population A I would augment has the potential to augment. So what government, since this panel is about government and governance, what governments need to start thinking about is how do I support my workforce in thinking about how to effectively leverage A I? And similarly, for enterprises, we actually did a research on A I and the BPO sector in the Philippines. And yes, there's a potential for massive displacement of the workforce. And so it's about how quickly can they go further up in the, in the packing order of sale sets, right. Um So, so it's a challenge, what you do see happening in Southeast Asia is still in that stage where we're figuring out adoption of A I. So we've got a bit of breathing space here for governments to think about how to effectively manage this. I think even beyond the comments you made, um thinking about how A I can supplement the existing workforce or better enable people to do even more beyond what they are currently doing. Um and the integral part of that is designing A I in a responsible and trustworthy way. If an enterprise is able to accurately see who is using A I as a company and how using it, they can actually maximize the benefits of things like turn of A I understanding and sharing prompts amongst employees. So that, you know, this is how my enterprise is using it. This is what's working for my employees. This is what's not working and putting your own guard rails around it, I think is a really important part to it, augmenting the workforce as opposed to replacing. So a lot of this is it creates communication challenges and responsibilities, not just for companies, employers but also for for our government to say, oh look, this is going to be net positive for all of us. There are going to be some winners and losers. But trust us, we're going to do what we can to try to adjust and adjudicate that get people the skills they need. Yeah, that's hard. That's hard to do. Well, I'm going to just add to say that. Well, I can't speak for Philippines and call center. I am quite involved in Singapore and legal sector and legal sector is one of those that expected, expected to be disrupted by especially generative A I. But I think it all falls on us wherever we are to actually take the bull by the horns, right? So the Academy of Law has articulated a four pronged approach to this. Number one, we actually need to just grapple with this and find ways to teach our lawyers, not only to use A I, but also to teach them differently. Because A I tools are going to make their way into tools that will help them become more productive and tools that their clients are going to use. We need to work with solution providers to get A I solutions into the market so that they are using it, right? And of course for us, right, we also want to make sure that the lawyers learn how to do things like pro engineering, right? And of course our own products and services, we need to also use A I to enhance our feature set. So it's really up to us where we are seated to try and get something done, right? And not just wait for the disruption to happen, but let's try and get ahead of that. Sounds like some, some thorny challenges that is unfortunately all the time that we have. But please join me in thanking our panelists for a really fascinating discussion.

Vivek LUTHRA, Senior Managing Director, Growth Markets and ANZ Data and AI Lead, Accenture  Debanjan SAHA, CEO, DataRobot Moderator: Ellie AUSTIN, FORTUNE

Brainstorm AI Singapore 2024: Unlocking AI To Boost The Bottom Line

You both work with companies across a range of sectors. Can you each highlight one or two industries that you think are really leading the way in terms of turning A I integration into revenue? But let's start with you. Sure, thanks. Good evening everyone. My name is Z. I'm based in Melbourne, Australia and I run our accenture's regional business for the A I I think the key sectors where we seeing uh you know, business leaders that are what we call, uh you know, strategic bets include datas like banking, insurance, health, those three would be the top three sectors. We see a lot of action. OK. You're kind of similar. There are four industries where we see a lot of traction. One is financial services including banking and insurance, health care, retail and manufacturing. There is a lot of actually work in manufacturing where they're using both generative and predict act. And are there any industries that are encountering specific challenges? Do you think? Uh look, I think uh surprisingly, the industries which are more regulated are actually using it a lot. You know, contrary to what uh I think uh you would feel uh actually, if you see whether it's health, it's regulated, financial services regulated. I think it's mostly uh manufacturing industries which are, uh if you think about the spectrum of A I, you know, classical A I, which is in terms of, you know, let's say predictive prescriptive modeling has been around for a while. I think as we kind of start to join generative A I to it, we see for example, manufacturing companies taking a little bit different approach, taking time to be able to, you know, get it. But we are already seeing some initial leaders in that sector. And when you say taking a different approach, what are they doing? That's different? Sure, you know, the kind uh the context is a little bit different. Uh asset heavy industries have different types of business uh conditions in which they need to work. Uh If you think about to get maximum value out of A I, you got to get your data sets ready. Uh And for asset intensive industries, the data sets could be very specific to the type of instrumentation they use, which is not the same as let's say financial services, for example. So I think those would be the kind of thought process which some of the clients we work with are going through. So bon if you're looking at kind of your portfolio of of clients, can you outline some of the specific obstacles that you think exist to unlocking the bottom line as, as the session is called, well, I mean, you know, I actually see uh three gaps which I think uh our customers are kind of uh dealing with one they call the value gap right there are, you know, a lot of excitement about A I but translating that into business outcome is not easy for various different reasons. Sometimes there are not enough uh tools and platforms that people can use. Sometimes they don't have the data, sometimes mapping the business problem into an A I problem is not easy for people. The second thing that I see, I call it confidence gap. There are a lot of people who are kicking the tires, they are doing prototyping, but they are not confident enough to take those applications A I applications or models of production because they are not sure about the accuracy of it. They are not sure about the safety and other risks that associated with those models. And third of course, is uh the expertise gap, right? I mean, this is kind of a huge area of focus for a lot of people. Uh There are not too many people who actually know how to build, for example, foundational models, right? But luckily you don't need that many people to build foundational model, what you really need people who can use these models to actually solve business problems. Even that I think requires lot of expertise and training. Other thing that I think is also going to happen, people are worried about is that do all of this, that A I is going to disrupt the skill mix in an enterprise. And people are worried about how, what impact it's going to have. For example, you know, various different jobs that people do today, what is going to happen to those jobs. And there are new jobs which are going to be created by A I which requires also skilling up the workforce. So what would be your advice to someone, a young person who is thinking about what to study at university? They want to thrive in this new A I landscape. What should they be doing? What, you know, I have one of those at home. My son just graduated high school and he's a computer science major in um University of California and he's kind of a little bit worried what is going to happen to the coding jobs, right? Because, you know, there are a lot of coding jobs which I think this A I copilots are going to do, but I do think it's also going to create new types of jobs and where I think it's going to be probably most lucrative and productive in terms of skilling up is that when you can marry the A I skill with, for example, a specific domain specific skill, right? For example, if you are an expert in biology, how to apply A I to do new medicines or personal medicine would be a huge thing. I think even, you know, some of the other areas, for example, education, I do think personalized education will be a big thing. So how do you use A I to create those new industries, new use cases, new businesses. That's where I think going to be the most action and that's what people need to learn. So that's what people should be thinking about as they, as they map out their careers. I'm interested in the confidence gap that you the expertise gap, Rebecca. That's something that you, that you've witnessed as well. And if so, are there any demographic specifics to that? Is it, is the, is the confidence gap more prevalent in some parts of the world than others? Is it women rather than men? Are there any specifics around, around those factors? Go ahead. Do you have, I think, look, the we are going through a large pivotal moment. Uh I think not having the right talent in the right places is a challenge which every business is facing right now. We just can't find enough uh you know, uh talent to go about doing what we call reinvention, thinking end to end. How do you really not apply this incrementally? But actually invent examples, you know, we worked uh uh maybe a little bit sidetracked but we worked with the food and beverages company uh in uh in uh you know, globally and there by using an A I driven solution for marketing content generation, we could actually build one year equal worth of content in about eight days. And that is phenomenal in terms of productivity enhancement. But at the center of it is how do you actually transform your workforce to be able to really scale this? One needs to think beyond use cases to business capability. You know, how do you think about a product launch versus a photoshoot in which you can kind of apply something and so on and so forth? So I think in terms of the question you asked, I think talent is difficult to find. Uh companies have to think about what does it mean from a workforce transformation perspective at accenture, we specifically have worked a lot on what we call what the future of work is going to be starting from there to be able to kind of think through. So if that is the future of work, what is the type of workforce you need and then be able to define which type of workers with what skill sets are needed and have structured curriculums on? How do you actually transform the work that kind of becomes quite key. And I can show you today, I mean, we are trying to get talent in every market we operate in talent is limited. So it's a combination of getting the right set of basic capabilities and training them internally in your firm. Actually moving people, I mean, conventional data versus modern data is a big area. For example, because if you don't get your data, right, how do you kind of really unlock value from A I and so on and so forth? So hopefully that gives ideas. So in terms of confidence gap, I think it depends on the risk level of the use cases that you are addressing, right? I think, you know, the EU A I act actually has done a pretty good job in defining those risk level. There are use cases where you probably should never use A I, right? For example, social justice and things like that. And there are use cases which are high risk, there are use cases, there are moderate risk, there are use cases which are low risk and it depends on the regulation that people have. And you know, one of the things I tell people that um the CIO S when you talk to them, they are concerned about three things that can A I create top line growth. Can A I save cost? And how do I stay out of jail and stay out of jail is where the confidence come into play. And if you talk to the cio s of regulated industries, that's where they are most worried about because it can not only create business risk, it can create reputational risk. Sometimes it can create many other type of risk. They can be on the wrong side of the law, for example. So what's your advice? To the leaders in this room on bridging that confidence gap, getting to the other side of it. Yeah, I think, you know, the first thing to do is to define the risk profile of the use cases that you're working on. And depending on the risk profile, there are different steps that you take. For example, what kind of testing you need to do, what kind of disclosures you need to have, what kind of moderation you need to apply to the results produced by A I? Right? So if you think through that in a systematic way, I think you can take care of the risk and you know, figure out where you can take it to production and what actions you need to take. If things go wrong, I'm gonna open it up to the floor for questions in a minute. Um I've just got a couple more for you both. Um So Banton, you've been a data robot, a CEO for about 18 months now, I I believe is that about it? Um And when you started the company was, it was in a bit of a rough spot. Think there just been quite a few layoffs, but you've really executed a turnaround. There's a new platform 9.0 you're rolling out various integrations. Can you talk to us a bit about the role that A I has played in that turnaround internally? Well, I mean, you know, when I joined the robot, um gene TB, I kind of was not the thing, right? So Gener TB, I happened, I remember I was actually in a Morgan Stanley conference in San Francisco. And everybody asked me about Gener TB I and the way back, the Uber driver squeezed me for half an hour on Gener TB, I, I knew that Gene TB I has arrived. And so Gener TB, I has played really a big cut, you know, part in turning around the company because there is so much of opportunity that I see with Generative Bi and what we have also done, we have been in A I business for a long time. We have been a P A company for about, you know, 12 years. But some of the things that we developed for predict bi applies quite well to generative Bi, especially when it comes to governance and observable and things like that. And what I have seen is that is the combination of predictive and generative bi the intersection of that you find the most value added use cases for businesses, right? And I think of this in terms of the lib brain and the right brain, the lib brain is your analytic part, the predictive A I and the right brain is your communicative part. And when you put the two together, that's when you get the iphone moment of A I which is what has happened, right? Go ahead. The, that's a question for you. So accenture has committed I think $3 billion to A I um how are you going to ensure that that investment changes the way that you interact with your client in a positive way, maybe accelerate the process. But you also maintain that high touch human element that accenture is so well known for. What's the balance that? Sure. And thanks for going there, I think it's a very committed investment from accenture. Our CEO could have announced that in 2023. Uh I think the key parts to that was how do we kind of really work with the industry and shape the industry's future part of that was how do we build industry specific solutions, functional specific solutions, uh build a network of what we call uh you know, uh Ja I studios across the globe, be able to really help our clients uh see what's the latest uh across the globe when they come to such a center. Uh The other part was uh you know, uh training our own workforce trying to kind of do that. So largely how do we keep humans at the at the center of this? If you look at most of the research accenture has published quite a bit, about half of the benefits come from automation to the bung point as you go from classical A I, which was all about data to generative A I which is so much about language when you put it together, what you can do in terms of automation, like, you know, autonomous processes that scale is amazing. Uh However, you know, as you kind of go towards, you know, putting this together rest half of the benefits are what we call augmentation based. So you got to start thinking not incrementally around processes, but you need to start thinking reinvention. How do you actually start to think in a new way, how you would do customer care and it's very, very different than looking at how you're doing it today versus how do you want to do in the future for doing augmentation? You got to start with putting humans to the center thinking about in this new world. What is the role which machine plays to kind of really support humans and still get to benefits and the benefits are not just automation, productivity benefits. The real benefits are around outcomes having different types of work, which is more interesting getting to what we call a 10 X impact in clients, business and so on and so forth. So I would think that's how we are looking at. How do we put humans at the center? Quick question before I open it up. The bon you mentioned before the iphone moment for a I, I'd love to know from both of you. What would be your personal iphone moment for a I maybe it's happened but maybe it hasn't. What would that look like for you? That light bulb, massive light bulb moment? That changes everything. Well, I mean, you know, it is when A I becomes pervasive in everything that we do, right? And I don't know if I'll call it iphone moment, but this is an example use case which kind of fascinated me for a while. Uh So we have customers who use classical A I to do various different kinds of forecasting. We have a grocery store customer, they own a lot of grocery stores in the US. And they use our models to predict, for example, how many mangoes they need to order in a particular store in California, right? And they have been doing that for every product in every store. Some various different days have different models and there are thousands of those more and what they have done with Gene Bi and which is quite fascinating. Now, they figured out how many excess mangoes they are going to have in a particular store. And they are creating personalized coupons that they are sending out to people that they know they are also creating, for example, very, very localized radio ads that they are putting in various different radio stations. For example, if they know that Indians like, you know, ripe mangoes and Mexicans like green mangoes, they will put a Hindi Language ad in a local news station and they are doing all of this thing automatically. So, you know, pretty much everything that we do probably can be made more intelligent with the use of A I and it's when it becomes pervasive and invisible, that's when you know that iphone moments will happen. So, no leftover, man, guys, that's very interesting. It's interesting. Look, I would kind of want to say that maybe last year it was mostly about experimentation and we work with businesses, but this year we are seeing businesses really put their heads together to be able to scale, to be able to scale. You need to think about, you know, the competencies across having a large language model is a small part of that which you use for doing whatever. There are normally seven things we see leaders who do this well, kind of bring a lot of things together to be able to scale this up. So think about being on cloud, that could be one thing. Second is getting your data ready, which we spoke about third, think about the enterprise grade A I solution. That's not just about the model on the model, you have adaptation on top of adaptation, you have consumption. So you've got to put that thing together, which becomes important. You also want to integrate that with your, you know, digital platforms, like your enterprise, the P platforms, the marketing platforms, the sales platform. So there are a lot of these things which come together to be truly be able to unlock the value and be able to scale up. And I think that to me is where we're seeing a big difference from last year to this year. Any questions for the V and a man? And if I raise your hand, if not, I will continue with one last question, I, I wanna finish on the subject of timeline. So you've just out laid out multiple steps. So that um if a, if a client comes to you and says, hey, we're about to invest a ton of money now in A I, this is our priority going forward. What's a realistic timeline for them to see meaningful returns? Sure. And, and let me take uh examples, right? Uh I think in the press uh uh accenture and National Australia Bank recently announced a journey we've been on with them for about 1.5 years now. Uh through this process, they focused on how do we invent uh uh using uh A I generative specifically in customer service and operational efficiencies. Think about 20 odd use cases, about eight of them being enterprise grade level. So that's about 1.5 years. What I want to say is I think the key thing is it is a multiyear journey, right? If you're thinking about reinvention, my encouragement to leaders I work with is to think beyond use cases to be able to go at scale, you got to move from use cases to business capability and that takes time. Take a lot of thinking, I would highly encourage all of us not to kind of really get bogged down by incrementally changing the process investing in your own workforce is going to be very important. And then thinking through all these seven conference we spoke about, but really it is a multi year journey. You know, you got to put, be thinking, how do you not going to do incremental changes to what you're doing today to be able to and one final tip for the people in this room, if they, you know, they in companies that they want to really unlock as we at the bottom line going forward. What's the one thing they should go into their offices tomorrow and start thinking about? Well, I think I do think use cases and returning value to business is important. I mean, you know, for J we are in the honeymoon period and that's not going to last very long, right? So it is important to pick the use cases which show near term value, but also have a long term plan so that you can show the value at scale, which I think is going to happen. But showing near term value and showing some return on investment, showing some successes, think it is very, very important. One final bit of advice pretty similar, I think lead with value. That's what we do with all clients. You've got to be really knowing what you're solving for from a business perspective. This is not a tech for tech. After decades of technology life cycle, you have something where the technology is adapting to how humans work. I think it's an exciting era if approached correctly. There is so much to be done in terms of unlocking value. Thank you so much for joining us. Thank you. Thank you very much.

Cha-Ly KOH, Founder and CEO, Urbanmetry Shaun KOO, CEO and Co-founder, H3 Zoom.AI Joe XIA, CEO, Jidu Moderator: Nicholas GORDON, Editor, Asia, FORTUNE; Co-chair, Fortune Brainstorm AI Singapore

Brainstorm AI Singapore 2024: How AI Is building the cities of the future

So Charlie first question to you. Um you know, a lot of places talked about being smart cities and the benefits from them are sometimes a little hard to recognize. Um there have been some pretty high profile failures as A I really developed as we get this new technology is the idea of the smart city now maybe coming into fruition. Thank you so much for having me today. And I think I'm really excited about the potential of A I primarily because it levels up the playing field. So for the longest time, smart city was pitched as this innovative, very, very effective way of governing cities, but it was afforded only by developed nations or developed cities, right? So the four processes to achieve a smart city, data collection, data analytics, decision making and action. For the longest time, the folks who can afford data collection was really developed cities like Singapore, many parts of China, Korea today, data collection can be done at a fraction of the cost using A I for developing cities, right? So that makes it bringing it forward in terms of leveling up the playing field for these cities. But I think to, I also wanted to curb some of enthusiasm, right? So like if you think about data collection that levels it up data analytics, great. You can now use some of the A I to do that. But let's talk about a little bit in the data decision making and action is that purely able to be done by A I in this part of the region. I think we are still quite far away from that because fundamentally cities are political and there are political parts of the decision making that unfortunately, we don't trust A I or I don't foresee that we trust A I to do that completely for us. At this point. I do want to get back to this point about politics and trust. But before I do, I want to bring in Sean, you know, as someone who is doing this data collection, could you tell us a bit more about A three zoom A I and what you're trying to do? Right? So at Atrium Zoom A I, um we see ourselves as the pent of the build environment. So we provide intelligent insights and actionable outcomes for property developers, fortune 500 property management companies and engineering consultants on the entire building condition assessments, essentially maximizing the life cycle value of these assets, right? So, so that's essentially like what we do, we work with a myriad of data sources, drones off the shelf cameras and we just ingest all these data sources to provide these outcomes in a very intuitive form. Um And I want to bring in Joe now, you know, autonomous driving is another one of these terms has been talked about for a long time. You're with G do, could you explain to us kind of what exactly is G do and how it relates to Baidu? It's A I initiatives with Ernie and the Apollo Robotaxis. Where does G do fit in this network? Uh Yeah, thanks for having me. Uh I think uh uh J uh first of all J is uh we're kind of gonna start up company. We're only three years. Yeah. Um So when we found it, um we're aiming to actually developing, we call uh smart EV, it's not ev it's a smart EV, right? Um So we call our product A Robocop. Um So, um so we have a joint uh shareholders, one is back to another, the GD. Um The goal for them to come in together is, you know, for the past 10 years that have, you know, that has been like invested over 100 billion um IMB on A I technology. Um So, so it's just the right time right now for uh for us to bring those A I technology, you used to be wrong on the cloud now, we can bring that to the cloud. So, so that's why it brings us here. Um And, and, and I think it's the right time because the computing power on the car side and both on the computing side on the car side are, are right now, is big enough to supporting those air to running. For example, um A 04 used to be like, you know, a very heavy on computing on GP US and you don't, we don't have a vehicle level um chip set in order to run that on consumer. Um you know, vehicle uh not until 2000, 2023 we have or X come from, you know, came from a media. So that's why we're doing that is uh uh uh we, we, we, we, we develop a car and we in great a be A I technology like board and uh you know, TB like, you know, A I um program and uh A AL four technology, but not exactly L four. But uh how are we gonna uh firstly, um you know, kind of downgrade the L four and having customers to be able to use a high level, you know, kind of atomic driving in the city. So that's the, the purpose of the goal of, of, of ju Yeah. All right. I want to stay on this politics point now, maybe first question to you, Charlie. Um You know, I, I've read some of your commentary on this and you, and you talk about things like housing, housing affordability, big political topic. I saw your linkedin Post about diesel subsidies in Malaysia. Another big political topic. How are you, how do you see data analytics and A I and data collection interacting with the political discussions in these cities, particularly in Asian cities. No, I think for the longest time because data was so hard to collect and things become a little bit more opaque. It was very hard for a data driven decision to enter into the conversation. I'm I'm saying enter the conversation because things are moving very quickly that by the time you collect this information, it's too late, right? But today, things are so real time that the conversations and the disparity between some decision making and the data that's available becomes a little bit hard to reconcile. So I think the availability of data gives a lot of transparency in terms of what are, what are people really being motivated by, by doing certain decisions, right? So the diesel point that I made on linkedin was that we are consciously thinking about how we are building a city very far away and that in itself is a conscious decision on how much we're spending to transport our logistics. So while we are solving it with ev while we're solving it with efficiency of A I, shouldn't. We also use this to plan in the future and reduce that today. So those are the part of using the data collection using A I will help us bring that point forward to be part of the conversation. Not really to dictate it, but at least participate in, there's also a risk that, you know, people can interpret the data, how they wish in ways that support certain conclusions. I mean, what's the risk of people then misinterpreting that data and thus worsening the political debate over a certain issue, I think uh that's always the case with any kind of data and however you um you collect them, right? And I think a lot of times people are really afraid of using data A I to do analytics because of biases, right? But I think it is even worse if you don't have the data available for people to not be able to actually drill down to to to even question the data. Data transparency is very, very important and the ability to allow general folks actually criticize that data, I think is step one. I want to go back to Sean now, you know, when, when drones were kind of the next big thing in tech, you know, they were cheap, they're vous they were everywhere and then we all got distracted by something else. Um You know, the two, you know, what else can drones help us do in terms of urban planning? And then how do you manage people's perception of those drones? They may worry about them being invasive, they may be worrying about personal privacy. So how do you chart that balance in terms of taking advantage of what drones can do? But also making sure urban populations aren't scared by fleets of drones like data on them, right? So, so, so that's actually a very good question. So, so obviously, um I think there's a lot of fixation on drones as being like the next big thing, but obviously, you know, through sort of conversations with partners, customers as well as industry um um stakeholders um at the end of the day, um what the the data, right that is being collected by drones, be it whether it's for construction monitoring, uh for asset inspection, for precision agriculture. Um what the industry and the customers really do care about are the insights, right? What are um actionable outcomes and what can um A I algorithms um that have been developed, you know, how, how can they actually interpret some of these unstructured data? Right? So, so I think the the the the perception definitely has changed over time. Um And, you know, from, you know, from my perspective, uh you know, being in, in, in, in the business for like close to 10 years now, um the sort of evolution of acquisition has also been extended to ground robotics, wall robotics. Um because, you know, the end client, what they really value and care about are outcomes that would drive the decision making processes, right? So I think in terms of governance around the data collection um and security aspect, I mean, there are, you know, existing guard rails um that um that are in place, right? So obviously there are algorithms out there that could, you know, mask uh faces windows, um you know, number plates, et cetera to sort of appease um and uh appease some of those like regulatory requirements. On the other hand, you know, there are um you know IO standards um that, you know, can be built right within um the uh the system level, right, in terms of data, encryption, sovereignty, etcetera, so that there are definitely like best practices out there, right? So, I mean, in short, I guess to drive home, the point is um you know, the the the the end the end outcome of being able to interpret the unstructured data to drive the business decisions are the true sort of a testament of what the industry and the clients are actually be looking over for. All right, one more politics question, then we'll move back to safer territory. Um You know, so Apollo's Robotaxis investors have been very excited about those, but I know that led to some controversy in Wuhan from people worried about how they will affect jobs. And you know, how do you, how should cities manage that transition as more stuff gets automated? How do they manage the human labor that, that gets affected from that new technology? Yeah. II I think uh uh just, just quickly answer the question. Um It's still far, it's still, you know, too early to think about that, right? Because uh um you know, uh Baidu has like started developing autonomous driving since 2013, um like 11 years ago. So the government under actually the supervised supervision of the government. Right. Um So, so, you know, started from a single vehicle, a few vehicles on a single road, you know, all the way to now, you know, uh just to be clear, not until last year. Um you know, um um um 40 L four robot taxi is not allowed in China to operate in the government just open that up last year. And, and, and then really with a small scale, not a large scale, talk about Wuhan Wuhan has over like 40,000, you know, taxi and, and, and the L four road taxi right now is only like 405 100. It's very small amount of road taxi. I think it's still under trial, the government, it's not, it's not there. It's not like the next day they're gonna be like thousands of hundreds of thousands of, you know, p in the road, right. It's, it's, it's still, you know, too early to, to worry about that. Right. I mean, still it's still a trial and, and, and, and most importantly, I think every time there's an evolution of technology it's going to open another window up, right? And you think about it years ago when we talk about, you know, um um you know, those robots on car manufacturing lines right. People thought like we're going to lose, lose our jobs, right? But turns out there's more jobs that are created because those robots, when they, those robots come in and, and, and it's actually increased the efficiency of the manufacture. And in that case, actually it makes the the sales of the car volume even bigger and it creates module, right? So, so I don't think we have to really worry about that at this stage. So automation will create more labor, I think ended up will create more, more jobs, you know, similarly on this labor point, I mean, um by, by replacing, you know, data collections, like people, you know, guys on the street with clickers, wiring the traffic with sensors. Um Yes, I mean, you will be, you'll be replacing jobs would normally go to people. How do you see that transition again? Also, especially in most of Asia where, you know, labor is probably still cheaper than capital, you know, how do you kind of get balance between labor and technology? So I think most of ASEAN is still thinking of the current situation, but A I is going to be built for the next 50 years, it's not going to be for today. So if you are thinking of your population as the next, of course, everybody needs a job. But is it for the job today or the job for the next 30 years? Do you want to be a guy standing there doing the quicker or the guy who's actually designing the A I system, who's counting the cars. So governments are actually in the position right now to make a call around the transitioning towards the higher value jobs in which you can actually create the value in the next 2030 years and not having to worry about stagnation, right? So stagnation is something that we should worry about today and taking the the step of actually using A I as augmented uh you know, help for you to, to do that is going to be a good system. I think um Sean, I do want to bring in the sustainability point, you know, making cities more climate friendly. Obviously, we're in a region of the world that's going to be affected by climate change. How will these technologies make cities more sustainable? Right. So, so obviously, um basically, the build environment is kind of like the bedrock of urban city centers. Um Obviously, you know, if you're using like robotics and you know, drone platforms, et cetera, um there are a lot of direct benefits um in terms of replacement of having to like ferry workers on site specifically in Southeast Asia, right? Because we still rely on a lot of low skilled foreign workers that are, you know, you know, essentially building the buildings and the city centers like from ground up, right? So obviously, in terms of carbon emissions, there are like indirect impacts on transportation that's one in terms of the data aspect of things by being able to plug into various data sources within building management systems. HV AC. There there is a potential to also reduce the carbon emissions from buildings, right? So 30% of carbon emissions today, you know, they come from buildings, right? So if we can reduce sort of like the footprint, in terms of how buildings function today through some of these insights, then it would, you know, empower architects and designers to design more sustainable buildings and essence, right? So, so that's how we, you know, how, how I see, you know, sustainability, I want to squeeze in a couple more quick questions for the time. We've got Joe, you were also previously before g do you were the co founder of Mobike, you know, bike sharing app, bi bike sharing service. What's the lesson from that experience that you think applies in terms of data and urban planning? Yeah, we actually um so, so people are just simply think Mobike is um bike sharing and we just dump bicycle on the road without thinking, right? And actually not, we're actually utilizing a lot of data. Um So um we actually work with a lot of transportation um research organization in China. So what we're doing is we're actually using the bicycle riding hotspot um like map called map and, and combine that with bus and taxi and those like, you know, ride hiring, you know, see how these, those three different types of transportation is gonna make up the whole transportation system for the, for the city. So we actually help the government to re replanning, replan their bus station because of the bike sharing, right? Because sometimes the last mile can be, can be covered by back sharing. And we actually use those data. Well, well, it's anonymous data, right? Um you know, work with, with government like combine those three system like data from different systems and and and use that for urban, you know, kind of planning, you know, to make people's like life in the city more easier. Yeah, super last word Charlie, how much is this like the game sim? So like the games in city, when you interact with the city, there will be an effect, right? I think the most of the time people forget that the people protest too. So I think if you remove, if you increase taxes and you start monitoring everybody, there is a danger as well of pushing them too far, they start going on to the streets, right? So I think as much as we love A I and data and data collection, we have to be wary that being in a city and being in a urban setting, it is a political thing and decision makers can use it to assist in collection of data and the data analytics. But beyond that, I think we do have to be worried about it. OK. Last time we got, please join me in taking all of our speakers.

Arvind JAIN, Co-founder and CEO, Glean Ahmed MAZHARI, President, Microsoft Asia Suthen Thomas PARADATHETH, CTO, Grab  Stephanie SY, Founder and CEO, Thinking Machines Moderator: Ayesha KHANNA, Co-Founder and CEO, Addo AI, Co-chair, Fortune Brainstorm AI Singapore

Brainstorm AI Singapore 2024: AI will have a $1 trillion impact in Southeast Asia, says Microsoft Asia president

Arvind JAIN, Co-founder and CEO, Glean Ahmed MAZHARI, President, Microsoft Asia Suthen Thomas PARADATHETH, CTO, Grab  Stephanie SY, Founder and CEO, Thinking Machines Moderator: Ayesha KHANNA, Co-Founder and CEO, Addo AI

Arvind JAIN, Glean, Ahmed MAZHARI, Microsoft Asia, Suthen Thomas PARADATHETH, Grab, Stephanie SY, Thinking Machines, Ayesha KHANNA,  Addo AI

Brainstorm AI Singapore 2024: How Quickly Can AI Be Scaled?

Jeremy Kahn, AI Editor, FORTUNE Moderator: Maiwen ZHANG, Executive Director, Executive Editor, FORTUNE China, Co-chair, Fortune Brainstorm AI Singapore


Brainstorm AI Singapore 2024: A Survival Guide to our Superpowered Future

So Jeremy, your book offers very dramatic predictions of A I's impact over the next decade. Suppose we can really travel through time uh to the year 2034 as a writer and thinker, what do you think will be the most underestimated change brought about by A I? Yeah. So I think there's several things, but I think we're vastly underestimating the extent to which this is going to change the way we kind of interact with the internet in a way that, you know, Google kind of changed the way we all search for information. I think the sort of A I chatbots and actually more agentic A I that's coming along is really going to change the way we interact with digital information in a way that's kind of very profound. It really is like the next platform shift. And I think we're in the future, not really going to do a lot of this searching ourselves. We're going to have these A I agents that kind of go out into the world and do our research for us, do our shopping for us, plan our trips. And I think that's going to sort of profoundly change the way companies potentially interact with customers and the way we all interact with each other in ways we haven't really grappled with. And I think a lot of companies aren't really grappling with. So I think that's one of the biggest changes if you project for decade, uh, that we'll see. I also think in all of us who are sort of knowledge workers, um, there's going to be kind of a copilot that works alongside us in our job and that's going to change sort of how we function in our, in our, in our daily jobs. And I think a lot of people aren't also sort of thinking too hard about that at the moment. I, uh, you also remind me that one of the, uh, feminists I listened to mention the topic called General Computer Control, which sounds harmless. But, uh, think about it, this could be a shortcut to a G I that A, I can understand what's on the screen, can, can take actions, interact with the cyberspace through a PC. So, if not properly regulated, this could be followed by things like feeding the public with disinformation, fake news, et cetera. So, is this like opening a new Pandora's Box? Yeah. Right. I do think so. That's the kind of agentic A I that I was talking about that. I do think it's coming. Um, but I think some of the risks you have to sort of look at, at how much of an additional risk you have from the technology. So if you talk about disinformation, for instance, we already have quite a lot of human generated disinformation out there. The question is if you have an agentic system, is it that much easier to produce something at scale and then also put it out into the world and kind of create a whole strategy for a whole disinformation campaign. I think what I worry about most with the sort of digital agents is actually there are some interesting issues about control. I worry actually about sort of scams and crime and this sort of thing because I think right now there's already systems where you can, you can tell chat GPT like what's the way I can make $10,000 in an hour and it'll tell you some ideas for doing that, but you have to then go ahead and implement that idea yourself. And so far when people have done these experiments, mostly Chat GPD recommends you run some sort of marketing thing where you're, you're then paid to market somebody else's product on the internet T shirts or something like that. But it could very well in the future be that, oh, you know, go make me $10,000 you know, in the next few hours and don't tell me, you know, I don't want to know how going to do it. I think you could very well get a system that says, oh, well, a great way to do that would be to run a phishing campaign for instance. And we'll just send a bunch of scam letters to people and tell them to transmit money to your bank account. And the question is if your, if your system goes off and does that like, first of all, who's liable for that, but also that's a risk that we really want to try to avoid. So I think the guardrails we have around agentic systems are going to be really important. So as who wrote, if we can try to govern A I, like what would, what we did with uh uh nuclear power, for example, uh what do you think our government and the international society can do now to prevent, you know, all this from happening? Yeah. Well, I do talk in the book a little bit about different analogies for governance including nuclear power. And I do think we need different levels of regulation. I think we should do some stuff at the international level that puts some basic limits on what this technology can do. I'm not as if you read the book, you'll see that I'm not super worried about the sort of existence risk scenario, but I'm worried enough that I think we should take some action to head that off. I think it makes sense to do that. I don't think that that should sort of crowd out what we're doing to deal with more near term risks. Such as disinformation, such as bias in A I systems. I think we need other regulations to deal with that and that could actually be done more at the national level. I do think every country is going to need a regulator that actually understands the technology has the power to kind of look at what technology companies are actually building and have some meaningful control over that and, and that we need some standards around what these controls should be. Um I do think there, there are risks here and we need to have a regulatory regime. I don't think you can just have a completely laissez Faire system. I think your book mentioned a lot of things actually are working. One of them is called uh constitutional. Yeah. Yeah, which is adopted by Amazon's uh large, large language model uh which is to give A I model written principles, a constitution to follow. So, so what it should be or maybe it's already written in this constitution. Yeah. Well, this is an interesting constitutional A I for those who don't know is something that was invented by the folks at Anthropic, which is one of the sort of leading uh LLM companies um founded by a group that, that broke off from open A I. Um And their idea was to give an A I system, a written set of principles that they, it should check every answer it's giving against and see if it conforms with these principles. This does seem to produce more robust guardrails than some other kinds of ways of trying to create guardrails. However, it's not foolproof, there are still ways around this. So I think we still need more research, but I think it's not a bad idea. Then the question becomes well, who principles are in this constitution? And that's a very fraught issue. And I think anthropic has done some interesting work on trying to survey the public on what do they think should be in this constitution? And they found, you know, two thirds of the people kind of agree on what should be in the constitution. The problem is that the other third vehemently disagree with the two thirds. So I think that's still, you know, we're going to have to come up with a way of resolving this. And I think it's, that's one of the reasons I think it's gonna be, there's gonna be geographic diversity among, uh, you know, chat bots. Um I think because we're going to need different guardrails for different places and different societies will think certain things are more appropriate than others, but they will be common area. Yeah, I think there will be commonalities and that's one of the things we could try to do at the international level is come up with, what are those commonalities? But is it, uh you know, uh constant intuitively, is it correct to say that maybe we need more powerful A I models in order to, uh, make sure that they, they will be safer because we, we simply need them to be that sophisticated to make it happen. Well, yeah, this is again an idea that the folks at Anthropic D Amede, who's the ceo of anthropic has this idea that you have to build 90% of the unsafe thing in order to do the research on how to make the system safe. I think that's a little bit dangerous way of thinking, but I do think there's a little bit something to it that we need to advance the system to a certain degree test, you know how we could control them. But it's also one of the reasons I think you need a regulator there sort of looking over the shoulder of the company saying, OK, well, at what point are we reaching a limit at which we shouldn't go any further? I also think there's some interesting research on if you built something that's relatively powerful, how do you distill the learnings from a large model down to a smaller model? And I think there's some interesting research has come out on that and I think some we may end up through building these larger models, create and more powerful models, create smaller models that are actually more controllable. But it will be because we've done the work on the large models. And the last shepherd, uh the, the, the very last ending of the book, uh actually brought me things some further last week, my kids uh and I visited the Mogao Grottos located in Northwest China. So uh sending interest from the Asian murals, mostly colorful paintings, religious writings on the wall. In the case, I was actually profoundly shaken by the fact that this is created. This was created by the very same human beings more than 1000 years ago. But think about that when our descendants saw writings, paintings, even music sheets, uh will they still be amazed knowing that this could actually be cations of machines? So with this fact, discouraged today's artists and writers, I hope not, I think it is a potential risk that people will stop appreciating, you know, human created objects and art. But I actually don't think it's a huge risk. I think there's plenty of room here still for human creativity. And if you look at sort of the most profound contributions, particularly, it's advancing the edge of any art, it's still going to be done by humans. It may be done at in collaboration with A I and using A I as a tool. But I think ultimately that to try to predict if you look at how a lot of these A I systems are trained, they're trained on past human works. And if you're trying to advance where art is going, it's very hard for these systems to actually make an advance on that goes beyond their training data. In fact, they really can't, in many cases do that. But a human can a human can potentially prompt, you know, use what the A I but then put a human touch on top of that, that does really advance the art form. And in the book, I talked to a number of artists who are trying to do this. There's a photographer I talked to who takes these big monumental of canvases panoramic photography. And then he applies an A I layer that provides some interesting visual effects. But then he does traditional photo editing. On top of that. I also talked to a writer who kind of takes the most bizarre outputs of systems and incorporates that into her writing and uses a kind of a foil for things she writes. So she has the A I in kind of conversation with her, which I think is interesting also as a way to think about how artists can still create human, written and human advanced art. So just as you wrote, we need co pilots, not autopilots. Yeah, that's right. I have to say we need co pilots and not auto. Yeah. So again, you wrote, I quote A I is actually bendable. So let's hope we can, we can uh skew the odds in favor of those more positive outcomes, not the other way around. Yeah. Thank you. Thank you. Thank you. Thank you so much. Thank you.

Josephine TEO, Minister for Digital Development and Information, Singapore Moderator: Clay CHANDLER, Executive Editor, Asia, FORTUNE; Co-chair, Fortune Brainstorm AI Singapore

Brainstorm AI Singapore 2024: Singapore tightens AI regulations on political content

Good afternoon, everyone. And Mr C, thank you so much for making time to join us. Glad to be here. We're really thrilled to have you and everyone in Singapore has been so helpful and supportive to us in helping us to understand uh how A I is being adopted uh here in Singapore and around the rest of the region and really uh appreciate that. You know, Jeremy mentioned that Singapore was one of the first countries to embrace uh and unveil a national uh A I strategy. And I wonder if you could for people who may not know about Singapore's strategy, say a little bit about that and, and perhaps how Singapore's new uh National A I strategy 2.0 is different than the first one. Well, uh thank you for that opportunity, you know, back in 2019 that we put forward the National A I strategy. It wasn't as though no work had been done. In fact, about nine months, even before the strategy was launched, we did put forward a set of principles for A I governance. So what was quite interesting for me is that even in those years, we thought about governance before we actually launched the strategy. Fast forward to 2023. When we put together our National A I strategy 2.0 the circumstances had changed quite a lot. By that time, generative A I had become something that was very accessible. And so a very important consideration is along with democratizing access to A I, how about the democratization of risks as well? You will find it being more pervasive, there will be many more settings that A I may be used or misused. So how does our strategy take into account both the democratizing access benefit as well as the risk, democratization of generative A I? So in any 2.0 I would say a couple of features stand out in terms of the infrastructure that can support an A I ecosystem. How much compute capacity do you make available, particularly for segments of the community such as the research community that may find it harder to access, compute that was one important feature. And of course, we also know that there is no good A I without good data. The question is, how do we harvest data in a responsible way, taking into account people's concerns about privacy, taking into account companies and enterprise concerns about data protection and foster an environment that has responsible as well as active use of data. And I think another very important aspect to go along with it is that in any new field, the pool of talent is always going to be a concern. It's not broad enough, it's not deep enough. So what measures can we put in place to build up the A I talent to support the healthy growth of an ecosystem? And if you wrap all of these together, I would also say that it has challenged us to think differently about what it means to achieve excellence in A I governance, which is not just the guard rails that you put in place, but also the promotional factors that can help this ecosystem grow. I think one of the things that has impressed me as I've learned about the Singapore A I strategy is that it seems to be very collaborative and there seems to be um a a pretty constructive relationship between, you know, as you mentioned, academics, uh people in government, people in businesses, um a lot of informal institutions and as someone who, you know, is an American and kind of keeping an eye on the American system of adopting A I. But a lot of time in China, I see the sort of huge polarization between those two systems where China has taken a very, you know, put the very dense regulatory framework in first and the US has almost an allergy to any suggestion of a regulation. And Singapore seems to have found this very interesting middle uh way uh where it doesn't have explicit regulatory framework, but it does have rules and systems and norms and, and kind of understanding which is, which is interesting. Do you anticipate that Singapore down the road will, will have more explicit A I specific regulation? Well, in the first place, there are certain kinds of A I generated content, for example, that we already have guard rails against. So for example, we have a lot of against fake news. How do you deal with the content if it is A I generated? Well, it turns out the law is technology agnostic. It doesn't matter whether it was a person who wrote up that fake news or used A I to generate that fake news and that already helps us to have some buffer. But there are also other areas of A I risks that need further effort specific again to A I generated content. What if the A I generated content was applied in the context of political elections? And this is something that already is observed in other countries where elections are taking place. I've said this before that we will have to look at what are the measures and what new laws need to be introduced. Um We are actively studying um the Korean, for example, put in place a law to ban political content that is generated using A I in the 90 days, that is the run up to their election. But in our case, we don't have a clear window over 90 days. So to what extent can we replicate a law like that, but I should just add that A I generation content is just one part of it. There are other aspects of risk that we also make available a set of principles that we'd like people to adhere to. So just as in 2019 before we had the strategy, we had the model A I governance framework this time round very soon after we published our updated national A I strategy, we also updated our model A I governance framework for generative A I. We've taken it one step further. We say here are the principles but how do you actually demonstrate that the principles has been adhered to? So we have a software tool kit and testing framework which we call A I verify. And that was very applicable in the context of classical A I traditional A I with generative A I. We created another project, we call it Moonshot. The idea is how do you test for the safety of A I systems with generative A I. So those are the practical ways in which we try to move forward. So it's an interesting combination of, of, as you say, guard rails and things to kind of encourage and promote innovation. You mentioned data which is a really interesting um uh component of this equation in the case of Singapore. So uh as I mentioned, when we chatted the other day, I was uh I, I spent some time trying to understand how a country uh like Singapore was, you know, 6 million people, less uh very small, limited land. Um what it does to get access to lots of data. Uh You will be familiar with Kai Fu Lee who speaks often at our events. He's got this book that argues that uh we're heading into this age of A I superpowers that are the the two big superpowers will be the US and China and they will be superpowers because they have access to the new oil of the Digital Age, which is data. So how does a small city state like Singapore compete in the age of A I superpowers and big data? So this point about the size of our population and the availability of data has been talked about before. Uh But I think you can look at it from a different perspective if you consider the size of our GDP and use it as an indication of the breadth and depth of activities that are taking place in Singapore. And that every single one of those activities generates a data point. Then actually maybe the data is not as small as we think it is, right? And in fact, there are also ways in which the training of A I models need not necessarily depend on a very large amount of data or the data being directly available to you. There are techniques in privacy enhancing technologies that will allow for example, Federated learning to take place. You may not be in possession of the data, but it doesn't prevent you from being able to extract the learning from that data. So investments in privacy enhancing technologies is one of the ways in which we can complement the need for data that also requires us to push ahead on things like cross border data flows, making it easy for companies that may have a multinational presence to be able to within their own organizations, be able to move data around. So those are the parallel efforts that we are doing in order to try and deal with the question of data. So my, my article is in the next issue of Fortune Magazine which will come out soon. I just went on the uh on the website today. But one of the things that uh I mentioned in there is that Singapore has been very proactive in thinking about how to use uh some of the assets that are unique that it does have to uh compensate for the fact that he doesn't have a huge population. Uh And, and uh there's the airport, which is one of the busiest airports in the world, the uh container port now, the second busiest container port uh in the world. You've got some of the the world's biggest banks uh right here. And then as you mentioned, there are all these multinational companies that are, that are coming in. So it makes for a very interesting ecosystem where there's a lot of technological sophistication and data, even though the population might be limited, the concept is very similar. Um whether it is for the aviation hub or the maritime center or for financial services, we've sort of position Singapore in the flow. You are in the flow of capital, you are in the flow of goods, you are in the flow of ideas, you are in the flow of people. And similarly, you can also be in the flow of data. That's how we're thinking about it. One of the bottlenecks on that, that's been a bit of a challenge for Singapore has of course been data centers and I wanted to ask you about that very quickly. Singapore has uh 70 plus data centers. That's an enormous number for its size. And yet this is a place that is resource constrained in many ways, land is scarce energy is expensive, there's no renewable energy to speak of and you know, it's hot and so running a data center in Singapore can be a daunting proposition. Uh Singapore had a moratorium uh that was imposed in 2019 because of concerns about uh data centers just consuming too much energy. Uh That's being lifted. There's a new uh blueprint in place. Uh And where do you see that going? Can Singapore continue to add data centers or will it have to start parking things next door in Johor? I'm really glad you brought up the fact that we're very hot. We can only ever give you a warm welcome in Singapore. You have to, you remember that, but let's go back to the question of data centers. It's not just the number of data centers that matter, it's the combined capacity. And if you look at our data center capacity relative to the size of our GDP, and if you compare it to say Japan or if you compare it to China, actually, we have way more. If you take Japan, for example, it has got a population that is maybe 20 times the size of Singapore GDP, maybe the 10 times the size of Singapore. But actually, if you look at the data center capacity, I don't think it is more than a 1.5 times more than Singapore. So you do that comparison across many other jurisdictions and you come to the conclusion that actually in Singapore, you already have one of the densest data center capacity, but it doesn't mean that we don't want to accommodate more growth because we do accept the fact that with A I being a more prominent technology undergirding our entire economy, you will certainly have additional demands being placed on data centers, more workloads need to be hosted in Singapore just so that the inferences can be done in the time that they need to be done. So some growth is going to be helpful. The question is how do we do it whilst also being able to fulfill our commitments to the net zero pathway as well as being responsible in sustainability. Tropical DC standards is one of them. So we introduce a new set of standards. If you're operating a data center in a tropical climate, do you need to operate it at the temperature? That is so low or actually even with one degree higher in temperature? You're OK. How do you do that? And then a more sustainable way to the future is to move towards greener diseases. This is either the operations, becoming more energy efficient or the source of energy that is used to power. The data centers will have to become greener. So those two pathways we think will give us some room as to how much it also depends on how quickly the technology evolves, actually, maybe very quickly if I could about um the the prospect for regional collaboration, as you sort of think about how to uh you know, project your, your weight in in terms of data. And, and A I uh Singapore is one of the most advanced uh A I adopters in this region, if not the most. Um But yet there's some data that has to be sovereign, has to stay on for how do you see the scope of uh you know, possibilities for collaboration with the rest of Southeast Asia and particularly uh Malaysia very much so because um ASEAN was one of the first to have a data management framework. It's been around for a few years. And on top of that data management framework, we've been able to put together a set of legal requirements that we can meet on both sides in order to facilitate the movement of data across the two jurisdictions. And this is not just between Singapore and Malaysia, it is pan a young. Um I would also say that there is a lot of interest and commitment to ensure that progressively we move towards more adherence to international standards in cross border data flows. And so not just Singapore, but several other countries are members of the US led global cross border privacy rules that was introduced some years ago. So it's not just a it's not just regional, it's not just trying to be really, really global and have a certain kind of multiple, multiple camps very much. So I think my colleagues and I amongst the Aan digital Ministers all recognize the importance of water. I would just the importance of data. What am I saying? What is of course also I would just like to say something in response to what Kai Fu has said about data being the new oil. One small problem with describing data as a new oil is that it gives the impression that data depletes that after you because once oil is used, you know, you can't reuse it. Data is not like that data can be reused many times over, it can be recombined and data is actually more useful when it is being reused in many different contexts. So when we think of data as a new oil, it is very valuable and that's what makes it look like the new oil, but it doesn't deplete, like oil will deplete. So we have to think of data in a different way. We do want to protect it are legitimate occasions where the data may be security sensitive. No country would want to share them. There will be data that will be enterprise, business sensitive and no right thinking business owner would want to share it and you need to find ways to protect them. But beyond that, there are a lot of benefits to being able to share the data and there are also increasingly more secure ways of doing so well. It's fascinating to see all the different things that Singapore is doing as I sort of mentioned in the article, uh Singapore is a great case study for other uh small and medium sized countries that are looking to kind of refine their strategy and to maximize the resources available to them. Uh Minister C Thank you so much for joining us today. Thank you. Appreciate it, really appreciate this opportunity. Thank you very much.

Presenter: Serena CESAREO, Senior Researcher, Tortoise Media

Brainstorm AI Singapore 2024: A Global Scorecard

Hello everyone. So the Global A I Index first launched in 2019 and has been updated four times since then. And it's essentially a big data led exercise that ultimately ranks over 60 countries based on artificial intelligence capacity. And before I go into the findings, uh I just want to say that since I've been working on it with brilliant uh data scientist, Joe White on the project, it's been just wonderful to see the amount of interest that this project gets from both academia and policymakers around the world who we know use our index to inform their national A I strategies and decision, which is amazing for us. And so in terms of its structure, we collect indicators uh data sorry, over 111 different indicators that we organize across three pillars and seven pillars that cover everything from research to development, talent, infrastructure, operating environment, government strategy and commercial ecosystem. In terms of the ranking uh for the latest iteration of the index, these are the top 10 nations and we see that the US ranks first on all pillars of the index and really leads by a strong margin compared to everyone else China comes second and it's the only country to get close to challenging the US in some areas. And Singapore comes third and Singapore ranks really well on all capita indicators. So all indicators um that kind of measure A I capacity uh in relation to the country's population and and uh economy size. And this is a good example for how a country that is perhaps smaller in terms of population and economy um can really outperform on A I and is uh a really strong and dynamic A I hub like Singapore is and talking about smaller economies, we are able to differentiate between intensity and scale as part of our index analysis. And we're able to do so by isolating per capita indicators. So like I said earlier, we don't just look at for example, the sheer number of data scientists in our country, but we also look at the number of data scientists in relation to the country's population. And by doing this, we see that uh Singapore tops the list when it comes to intensity. Uh but there are other countries like Israel, Switzerland, Sweden, Korea, Finland, and a few others that really outperform on A I. Despite being smaller countries, when we look at scale, obviously the US and China lead the rankings and I know I have to wrap up soon. But given that the 2024 edition of the index will launch in less than two months. I thought I'd give some preliminary findings from the the new iteration. And I've selected a couple of indicators that I thought were most interesting uh also make it regionally relevant to Asia given the opportunity today. And so this is the first one and this is from our com uh commercial ecosystem sub pillar. Um And this is looking at private investment in A I and we can see that Asia excluding China makes up 2.6% of global private investment in A I, which is very much dominated by the US as we can see here, which I think uh makes up around 66%. Uh then China is around 9%. Uh And the rest of Asia is 2.6% and it has been declining uh slightly over the past couple of years, whereas China and the US have been growing and so zooming in on uh Asia. Um this is uh a private investment in A I in the region. Again, excluding China is dominated by Singapore, Korea and India and Japan to a lesser extent. Um although yes, it has been declining over the past couple of years as you can see more clearly here. And then lastly, in terms of talent, this is a new indicator that we have in the index which is looking at the retention rate. Uh so looking at talent retention and not just talent growth, um and we're able to access data from a company called Dey Uh and this is showing that Singapore's capacity for retaining talent in A I has remained relatively stable over the past few years compared to um I guess other countries in the region like Korea, Japan and Taiwan, although its overall capacity, despite being quite high, around 70% is a little bit lower than that of other countries in the region which gets gets closer to 8 80%. Sorry. And so this is it for me, stay tuned for the September 2024 launch with more and improved indicators with an expanded list of countries as well. Um And yeah, look out for the new index on the Tortas media website and I'll hand over that to Ellie. Thank you so much.

Serena CESAREO, Senior Researcher, Tortoise Media Simon CHESTERMAN, Vice Provost, National University of Singapore and Senior Director of AI Governance, AI Singapore Shameek KUNDU, Member of Singapore Government AI Advisory Council Andrea PHUA, Senior Director, National AI Group, Director, Digital Economy Office, Ministry of Digital Development and Information, Singapore Moderator: Ellie AUSTIN, Deputy Editorial Director, FORTUNE

Brainstorm AI Singapore 2024: Strategies for Small And Medium-Sized Economies

Serena CESAREO, Tortoise Media, Simon CHESTERMAN, National University of Singapore and Senior Director of AI Governance, AI Singapore, Shameek KUNDU, Member of Singapore Government AI Advisory Council, Andrea PHUA, Ministry of Digital Development and Information, Singapore

Serena CESAREO, Tortoise Media, Simon CHESTERMAN, AI Singapore, Shameek KUNDU, Member of Singapore Government AI Advisory Council, Andrea PHUA, Ministry of Digital Development and Information, Singapore

Serena CESAREO, Tortoise Media Simon CHESTERMAN, AI Singapore Shameek KUNDU, Member of Singapore Government AI Advisory Council Andrea PHUA, Ministry of Digital Development and Information, Singapore

Kisson LIN, Co-founder & Chief Operating Officer, Mindverse AI Sambit SAHU, Senior Vice-President, Silicon Design, Ola Krutrim Leslie TEO, Senior Director, AI Singapore, Lead, SEA-LION Caroline YAP, Managing Director - Global AI Business and Applied Engineering, Google Cloud  Moderator: Matt HEIMER, Executive Editor for Features, FORTUNE; Co-chair, Fortune Brainstorm AI Singapore

Brainstorm AI Singapore 2024: How AI can bridge Asia’s linguistic and cultural divides

Kisson LIN, Mindverse AI, Sambit SAHU, Silicon Design, Ola Krutrim Leslie TEO, AI Singapore, Lead, SEA-LION, Caroline YAP, Google Cloud, Matt HEIMER, FORTUNE

Christine Yuan HUANG, Founder and CEO, Quantum Life Alex ZHAVORONKOV, Founder and CEO, Insilico Medicine Moderator: Jeremy KAHN, FORTUNE

Brainstorm AI Singapore 2024: The 1,000-Year-Old Human

So the title of this panel is deliberately a bit provocative, but there are several ways in which A I you could think about potentially extending human human lifespans. And one is through better treatment of disease. That's one way of looking at it. Another might be well through wellness, you know, sort of helping us live healthier lives. Another might be to actually try to treat aging specifically. So taking some sort of compound that would reverse the aging, the aging process. And lastly, and probably most sort of science sci fi would be the idea that somehow we're going to through brain computer interfaces or some future technology kind of merge with, with an artificial intelligence. And while our, you know, we might be able to extend our consciousness, even if our physical bodies don't, don't make it. Um I wanna, I wanna ask our, our guests because they're um they are experts in this area where they think the most impact from A I will be Christine. Let's start with, with you and then, and then Alex. Thank you. Thank you so much. Thank you, Fortune. Thank you Jeremy for the wonderful panel. Um I'm Christine. I'm from quantum life. And the quantum life is actually doing quantum enhanced artificial intelligence for precision medicine and for integrating human beings longitudinal data to predict more and more accuracy plans for human um health span. And of course, maybe later on its lifespan. So uh we are so excited for this new technology merging up and we are we have already seen for human life. And for the past like 30 years, we have already had a lot of changes in um Medicare in healthcare in pharmaceutical air driven pharmaceutical development. But the question is for human beings, we always think about birth, aging, disease and death. We are in Singapore. Some people are talking in uh in Chinese, we call it Shen Laing, right? We never think about we previously, we are all focusing on disease which is a disease and death. But how about aging before disease? That is how we wanted to introduce A I to improve before disease. And we have already seen during the past 10 years. Liquid biopsy, which is highly um based on the bos metics being there, which is make great progress in cancer prediction. And for disease prevention, which means a shift from the disease care before had based on the autonomous diagnostic. Before this, we had Q prevention. And next question that is going to be how we develop and interventions through the process. That is how A I will be introduced and we are going to move forward So this is on the idea of prevention as a way to forestall disease and to improve lifespan. Alex. I think you may have a, well, you may agree with some of that, but I'm curious what your view is on this. I think you might have a slightly different perspective on where you think the biggest impact will ultimately be on longevity. Sure. Uh So I um I wear many hats and one hat is in Silicon medicine, CEO and another hat. Um I co-founded a conference called Aging Research uh for Drug Discovery. Uh It's an annual event in Copenhagen uh every August uh for five days. A lot of people come to this conference and uh we convene and talk about aging. So it's probably the largest event right now. Uh And when I started the conference, um most of the scientists were talking about diet exercise, sleep, behavioral medication, right? And then of course, you know, about 30% of the delegates at that conference are also going after very, very visionary things that are way beyond that. But at the end of the day, um it always goes back to diet, exercise, sleep, uh and the hero medications. Uh And uh what I'm afraid of the most is that in the next 10 years, I'm already 46. Uh So kind of for me, it's all like you, you, you, you look into the future. It's all down hill, right? Sorry for both males and females at that time. So, whatever I do, it has to have more impact than just diet, exercise, sleep. So I think at most conferences, we need to start thinking about, you know, diet exercise, sleep prevention, diagnostics should be the baseline, right. So let's do what your mother told you. Just let's stop talking about us, do it, right? Some people are still smoking, right. Um, so we need to have something much more substantial, a big breakthrough cha GP T moment in longevity in order to get people excited and get them, steer them away from, you know, diet, exercise, sleep, behavioral medications, do what your mother told you because we know that works. It will work to some extent. But billions of people live their lives of them have optimized for diet, exercise, sleep and they still aged, got diseases and died. So we need to go radically to the next level. So far from what I've seen, nothing works on humans, there is nothing clinically proven to. No drug is clinically proven to demonstrate age reversal in human beings. So we are still trying to prove that some of the drugs that are known for 3040 years may have some impact even though they work in mice really well. So like creamy and for example, 15 to 20% consistent life extension in mice. And now people are doing combos looking at different ways to different molecules to go after aging. But I think that we need to have a fundamental breakthrough in order to live longer. And of course, things like um you know, brain clear interfaces very promising. So there you are not only maybe reducing the rate of loss during aging, you might actually augment your function, like think about glasses, right? Very easy innovation, but it restored a useful function. Plus nowadays, with 3D glasses, um you might be able to with a reality, you might gain new functions. So I think we need to go and look at the next level. I think that the next level is going to be drugs that target aging and disease. At the same time, if you are just looking for using A I to discover drugs that only go after aging, I think you won't be able to develop it into a drug for many reasons. One is just basic economics, you won't be able to reimburse it. Most people understand how difficult and costly it is to develop a drug. And also you have to be committed for 10 years or longer, right? So I run in silica for 10 years. My most leading drug is in phase two clinical trials. So in, as they say, in the Middle East, uh if we are successful with phase two and we go and you know, target the disease uh and show the proof of concept in maybe like 34 years, we will be able to show that it works on aging. Hopefully because that's how we discovered that using aging research. Uh So you have to be very committed for a very long time to demonstrate even a little bit of, of, of efficacy. Alex. Can you talk a little bit about how A I is, is maybe speeding up some of that process? And I know you're getting this but um have some of the most number of A I discovered drugs currently in various phases of clinical trials including in phase two, a drug for idiopathic pulmonary fibrosis, which is a drug that's targeting both the disease but also has interesting implications for aging. So, can you talk a little bit about first of all, like how A I can accelerate that process that 10 year drug discovery process? But also this dual strategy of going after things that both target unknown disease but may also have implication for age reversal. Sure. So first I need to kind of tell you and that's probably useful for all of you. A story of how drugs are discovered and developed. So, drug discovery and development is split in two phases. Basically drug discovery where you discover the drugs and ensure that you complete all the preclinical experiments and on development, it's human clinical trials. Phase one safety, phase two efficacy, phase three, safety and efficacy. That part cannot be dramatically expedited because you have to move with the speed of traffic, it's very regulated. Uh but the part where you discover drugs can be substantially accelerated. Um And so far, we've demonstrated at least 50% increase for the slowest of our drugs. In some cases, we've demonstrated that we can shave off 70% of the time and maybe 90% of the cost of drug discovery using A I. Um And drug discovery is split into three phases. One is formulating the mechanism of disease. So you need to understand why disease happens. So usually for Alzheimer's, for example, we still don't know why it happens and what's driving it. Then after you formulated the mechanism of disease, you need to understand which proteins or other features in biology are driving those diseases, we call them protein targets. So you identify protein targets that may work in a disease. Then you design a small molecule, a drug or a biologic, an antibody that disables that kind of criminal that you've identified that is driving the disease and that's a completely different kind of type of work that you need to do. Usually that's very two different separated areas in the pharmaceutical company. And then once you've designed the molecule, you need to conduct safety studies. So you need to go into animals and demonstrate preferably in more than one animal that the drug. Well, first of all works, so it's effective and it's also safe. So you call it I and enabling studies, safety studies and can start humans. So in our case, we use kind of life models, the models that understands human biology from cradle to the grave and also understand diseases to pick those protein targets that are implicated on aging and disease at the same time. So there are few. And then we have another form of generative A I that designs small molecules with the desired properties. Instead of looking for a needle in a haystack, you generate a bunch of perfect needles and then you synthesized test, go through the clinic, preclinical trials, go into clinical trials and then we either take it ourselves or we sell those molecules to other parties. So we actually sold a couple of drugs very successfully. And you know, last year we did $1.5 billion in total deal value selling those drugs. So those A I generated drugs that are faster and cheaper to make are actually of higher quality. So people are willing to pay a lot of money for it, right? Christine, you were telling me as we were coming in that you're also sort of looking at this idea of creating a life model using A I but maybe also to pick up on indicators of disease earlier and then you can have earlier intervention is sort of your idea of how you could extend life. Can you talk a little bit more about, about that and what you're doing? And then I want to go to questions from the audience. So please think of your questions. Hopefully we'll get one or two in. Yeah, exactly. Thank you Jeremy and, and just as I have mentioned, um because for drug discovery, it's a very, very long journey. And currently we also have the drug discovery from the preclinical stage. And Ino did amazing work driving forward the A I drug discovery to the to the clinical stage. But the question is we still need to facing the clinical stage, right? We still need to take like 10 years to get through all the regulatory approval. I have to say for bringing in new medicine or bringing in effective drugs or cures or intervention to our life. It's not only about pharmaceutical, it's not only about drugs, it's also about the one regulatory. Secondly, what is a personalized red drug for ourselves? So during the past 30 years and during the medical, if we leave our Western medicine is true, like to the past 200 years, we will figure it out. Ok. So actually the medicine drug, the drug development, the system is all because of based on one fundamental for medicine, which is, which is autonomy and that is a fundamental curriculum in medical school. Like I'm training as a medical doctor. So that is the the first year of curriculum. But during the past 10 years, we already shift our diagnostic window before the traditional autonomy because you think about it, how we diagnose the disease, we do autonomy and we have our brain, our liver, our stomach, right? And now what has happened during the past 30 years, the next generation sequencing has already shown up and a lot of molecular information has already been discovered. And together with the BOS Matic, which is the A I for the past 10 years, we have already developed several pre diagnostic tools like Guri developed the cancer prevention. OK. So the next error going to be, think about it, we have already shift a little bit the cancer diagnostic stage from the late stage like two and 32, the four stage zero. And what is the cure for the stage zero? And we need to develop a cure for the stage zero. And that is how quantum life wanted to discover from stage zero. And we compare that is not only disease we compare which is younger, healthy group and older group. And we discover what is the red intervention, what is the red target before any disease happening? And we develop a cure for the stage zero. So that is how we can make jobs for extending the life span and health span. Then think about it if we have this kind of interventions already, maybe drugs, maybe some other treatment like medical device can already be a treatment to the C zero. And what is the next step? The next step? That is how we can compare. Where is the right? Ah If we are talking about the precision Medicine, where is the database, how we can have the database, how we can make the individual personalize the medicine for ourselves. So we also need to have the database for ourselves. That is, you have to, you have to record a lot of data about each individual person in order to deliver the precision medicine because the data you have on the whole population may not be actually be relevant. That's interesting. Let's take some questions from the audience, at least let's get one in while we have time. Who has a question? Please raise your hand. We'll get a mic to you. There's I see there's a woman at the table over here. If we can get a mic to her, please, there's someone running over to you and if you could please stand up and state your name, that would be great. And where are you from? My name is Mia. So I'm Global Cio for it from G LP. So I have a question around. So the A I, the role A I plays in medical and where do you see the obstacles of further utilization of A I? Either in um this is not just precision medicine but diagnosis of a disease but also in a longer term health care really care about people. Is that a compliance issue or government regulatory or other obstacles? So I guess what, how you know, how do you make sure that A I is actually helping us bring out better treatment, maybe assisting doctors in a better way and what are some of the hurdles for that, whether they're regulatory or not? I know Alex was talking about a number of regulatory hurdles on that he's facing in terms of just getting, getting treatments out. But I'm curious on this wider question, do you have a view on, you know how a I might assist doctors in picking the best treatments for individuals and then maybe Christine can answer. So sure, actually, I will give the time to Christine on that because she is working on that area more than I do. Um So on the population level, how do you um use A I for health care? Uh That is um the dream, it's actually the actually the dream for human beings and for the current pub public health system, that is to say that is how we wanted to bringing Medicare beneficial to human beings and through medical doctor system as soon as possible of. So the, so the first of all A I for diagnostics, we can exactly enhance the diagnostic errors of the make medical doctor, radiologist to, to, to, to have a better, have a better diagnostics. And also for me, doctors can select a better drug and bringing the drug repro because some of the interventions may not need a new drug. We need to figure out how is the right drug for the right person? Maybe that the drug is for traditional drugs like Metformin, Metformin is very cheap and Metformin. Currently, we also see the um effect on extending human lifespan. Basically, it has been, but it's very cheap. So uh we are considering if we can use it and bring it to the healthcare system, I have a question on this, which is sort of, I feel like there's a lot of these medical techniques that are sort of have have been invented. They, they there have been a number of breakthroughs yet, the distribution of that is a problem. So we have this big inequality issue depending on where you live or even which doctor you happen to see your doctor might know about this breakthrough treatment, might be able to prescribe. It might do genetic testing, might not do genetic testing. I feel like there's this great disparity in the health care we receive just on some of these very arbitrary factors about where you live and what doctor you happen to have. I don't know if you can say anything about that Christine and maybe A I can help sort of bring the general standard of care up to a certain level. Exactly. Exactly. So, yeah, I think that A I is the ultimate um uh democratizer, enabler, and unifier, and equalizer. So now with C GP T, everyone here in the room can uh you know, can have more access to knowledge than some of the best doctors just five years ago. So you can take a picture of uh you know, your wound for example, and let it annotate it. And I hope it doesn't hallucinate an answer. Well, yeah, it has a problem, right. But it has a very good demo of what is possible because you give it a little bit more reinforcement learning. And if you don't need a snappy response, you can actually, you know, have a multi agent system that will try to solve your question in maybe a little bit more time than it doesn't hallucinate where you eliminate some of the areas where it could hallucinate. Um And it can answer very accurately. So it shows you that with technology, we actually not only uh equalize some of the professions, but we also make them very accessible to everybody democratize and we gain superpowers. And I think that we need to use this technology right now as much as possible to identify interventions that can target aging at a core and make this technology available to everybody. That's why we are very focused on small molecule drugs because small molecules, unlike biologics, they are super cheap to make. Once they go off patent, they are available to everybody on the planet for peanuts and they are also very easy to store to transport and people are very used to pills, right? Rather than injecting expensive biological. There's lots of potential for A I to democratize access to health and access to hopefully these new medicines, we're out of time everybody. But thank you very much for listening. And thank you to our panelists, Alex and Christine. Thank you so much. Thank you, Singapore. Thank you. Thank you so much.

Satyen SANGANI, Co-founder and CEO, Alation Geraldine WONG, Group Chief Data Officer, GXS Bank Moderator: Nicholas GORDON, Editor, Asia, FORTUNE; Co-chair, Fortune Brainstorm AI Singapore

Brainstorm AI Singapore 2024: How To Deliver AI-Ready Data

Torrey SMITH, Co-founder and CEO, Endiatx Moderator: Ellie AUSTIN, FORTUNE

Brainstorm AI Singapore 2024: Swallow This! The Transformative Power Of AI-Enabled Micro-Robotics In the Human Body

Brainstorm AI 2024: The Transformative Power Of AI-Enabled Micro-Robotics In the Human Body

So as I said, we're talking about micro robots here that are ingested into the body that in itself poses a lot of questions. It sounds a bit like a science fiction film, which is both a source of fascination. But also I think for many people probably in here a source of skepticism. So if we could start off by you explaining to me briefly why you think pill bot is more effective than the traditional endoscopy? Well, thank you. Our hope is that with little tiny robots that can swim around in the body, you can actually supplant or augment the existing tech that's out there. I'd, I'd like to think that pill bot could be a little scout to go out and find illness out in the world and then bring that illness into a hospital to treat properly. Ok. Ok. So it's, it's speeding up the process that it's making it more efficient. Um Looking at your website, the pil bot team, it's made up of a lot of engineers. Now, some people might expect there to be doctors, scientists on your core team. Can you explain to me why that isn't the case and where the medical expertise comes from in this process. Well, so besides a group of almost 20 deep tech engineers, we actually have leading gastroenterologists from all over the planet that have joined the company and many of them have even stepped up and invested and put their, put their names on the line. And are they part of the team or are they an advisory panel? What's the structure? Well, we're very proud to count Vivek Kumari, a leading gastroenterologist as one of our co founders and a very close friend. OK. OK. So some people here in the room are going, I would imagine the draw comparisons to something like Elizabeth Holmes and the an of scandal in that this is an outlandish idea that promises to transform the healthcare industry. Some might say it's too good to be true and we're going to see the technology demonstrated in a minute. I mentioned that you're in the process of seeking FDA approval. How else are you going to legitimized pillow within the scientific and the medical community to silence those skeptics? Well, the key here is just uh putting a robot where our mouth is, right? If we can create a technology that lets a doctor have a look around inside you to the point where it's maybe a little bit more inexpensive or a little bit more accessible. And if doctors start to say we'd like to use this in our actual patients, then we're starting to do our job. But until we get to the point where doctors are demanding to use pill bot, you know, it really is just a R and D project, but we're excited to show it off right here. So you can kind of start to judge for yourself. So it's word of mouth. Obviously, that's the FDA component is that also going to be peer reviewed research. How does that side of things work? Actually, that's been one of the things that's helping us most of all uh attract brilliant doctors is uh to be honest, uh everyone wants to publish on til bot. It's an exciting new platform that can open up all all sorts of new research avenues. And so people are actually realizing that uh you know, joining the team and, and getting a hold of the robots could actually help them in their, their academic career as well. And are people publishing? I know PBO? Yeah, it hasn't featured hugely in the mainstream press. But if we're looking at the scholarly academic side of things, kind of what level of publication has that been so far around Pil Bot, the biggest publication so far that's coming is that we have officially done our first patients uh in Auckland, New Zealand uh in our IRB uh clinical trials. And we're actually starting to see Pil bot do some clinical work, but of course, that's just uh under, under under active progress right now, trial gets complete, will basically bring all that data together and we can think about publishing. Ok. So nascent, but you're looking to have more data in the future. For sure. My final question before this part of the session is so pilot or your, your company has raised $7.5 million I believe in funding since it was founded in 2019. And it's worth noting that 1.5 million of that dollars, um 1.5 million of those dollars comes from the Singapore based Virge Tech Health Fund. Um 7.5 million. It's a good amount. It's not a huge amount. And I imagine that this is deeply expensive technology to develop. So where's the rest of the money going to come from? Well, we're obviously excited to continue that adventure. You can always do things with new amounts of funding. But the thing that we're really proud of is with the support of Verge Health tech funds here in Singapore and the support of some of the most passionate deep tech engineers in Silicon Valley. And, and actually beyond, we've been able to make some magic happen using pretty limited resources. But when we show you the robot today, you'll see that it's not perfect that you'll definitely see opportunities for it to improve. And that's kind of where the next rounds of funding might be able to help us. Ok. So you're going to continue pitching right on that note, I know you've got a couple of slides to show us to uh to explain in a bit more depth how the robot works. So take it away. Yeah. So let's just briefly look at a robot and we'll start to ask ourselves, you know, if we go back in time, you know, the way that you would get data uh from someone's health inside the human body, um was very difficult, you could cut into the body, you could slide a tube into the body. But in a hospital and we're asking ourselves, could we potentially have microscopic robots travel throughout your body? Not only finding out what's wrong, but maybe actually starting to fix it. So that's the adventure we'd like to go on today and we'll have to acknowledge that it's a difficult challenge. But where we're going to begin is the human stomach. Because in the human stomach, we have this unique opportunity for a robot that is maybe not microscopic. This is the size of your fingertip. But pill bot contains a live video, three dimensional motion with pump jets. If you're willing to drink some water, you can get some motion um and the ability to control it in XY and Z. And so the goal here is to basically show there's a part of the body that we can actually send a robot to right now. The stomach, we can probably do an amazing block of work there. Then where do we go next? And so you'll see right here is kind of where we are in the current process of miniaturization. I want these to be rice brain size. I want to do brain surgery with it. Um So looking inside till bot, you'll see a pretty elegant combination of pump jet motors, really high powered lithium batteries that we work really carefully to make safe live video, right? A lot of the stuff you would see in a drone. But the reason that we're doing these robots is that they give us the data, right? Data is how we understand what's wrong data is how we train up A I. And from here, there's a bright future for this. But you know, let's start to talk about what we can actually do with uh one of these. So OK, this, this is pill bot and I need to turn this thing on and if I'm gonna turn this thing on, I'm gonna need my chief engineer Quentin to come out here, Quentin. Let's bring this thing to life. OK. So, and Tori and Quentin, I need you to talk us through what you're doing here at. So Quinton is using an optical signal, a bright light to wake til bot up, it's been sleeping. And so now you'll see til bot um start to uh swim around. He just got some water into the thruster lumen. So we don't get vapor locked. And what we have here is basically a swimming little drone. And can I just highlight how Quentin is controlling it. He's using what looks like an Xbox controller. Yeah, we, we find that a lot of younger gastroenterologists really understand the value proposition. Um, but basically what we're doing here is we're saying, let's make a swimming robot. Let's call it a little eyeball. Let's drink some water. Let's swallow the robot. And let's take a procedure that would be done in a hospital after several physical hospital visits. Let's take that whole adventure and let's digitize it. Let's turn it into a telemedicine, call it home. Let's let a I do some of the boring work for us. So this is about accessibility. How much does a pill boat cost? So cost of goods uh can often hover around 35 to 50 bucks, right? It's a relatively inexpensive physical article. All of the value is in the hard work that miniaturized it and in the data that comes off of the robot. But if we can't make it inexpensive to build, then there's no way we're going to get it out into the world and actually go find problems. What do you expect it to cost when it comes to market? Um In the US market, it'll be a couple of $100 just like a normal pill camera. But if we're going to take it to Asia and beyond, um honestly, I envision a future where these would cost maybe 10 or $15 or maybe even less, but you'd have hundreds of millions of people accessing this kind of care. But the amazing thing is that while we're doing that, the health data that would come off a platform like that would be incredibly valuable. OK? Um So Tori, I believe you are not going to swallow pill bot. I, you know, honestly, we came all this way. I think we might as well might as well go ahead and swallow. It would be rude not to. Ok. So you're, you're gulping down a glass of water to prepare your stomach. I imagine I need a little fish tank in my tongue. You need a fish tank in your stomach. Ok? And for context, Tori told me earlier, this is the 45th pill box that he has swallowed in his lifetime to date. It's, it's a lot of fun. We really enjoy this process. Ok? Perfect. And for those who can't see, I was like, oh wow. More. Ok. I would say the pill bot looks like the size of a pretty large multivitamin. Um It's, it's not small, it's not a small robot. In fact, it's probably just about the biggest thing that someone could ever reasonably swallow. And even maybe half of people might not be able to do it. But the thing that bugs me is that we transact as much as $55 billion every year in the old way of how we do endoscopies. But that only reaches 1% of stomachs and So we're, we're really not reaching people where we need to reach them. And what do you think? Should we uh go ahead, we get you guys. OK. All right. So Quinton is gonna hand me this guy and uh let's uh let's do it. I'll get it nice and wet and uh hopefully let's get a live video up on the screen. Here we see. Oh, here we go. There we go. OK. So you can see. Oh, and I need the uh antenna a little bit closer to my tummy. There we go. So hopefully you can see me. You can all see yourselves. And now let's go ahead and swallow pills out. All right. So let's have a look at my stomach. I've been starving all day. Ok? I can whip it around a little bit and then the hope is that we'll sort of break into the bottom of the stomach and you'll kind of see what's going on. There we go. So, the idea here is we want to make sure that you can quickly and easily give a gastroenterologist a live view inside the stomach and the ability to move around. I want this to be like the tip of an endoscope, but just in virtual form. Ok. Um Here we go, we're traveling through your intestine. Um Tori if how are you feeling, first of all? So sometimes I'll feel it when I swallow it. But honestly, once, once I swallow it, that's OK. I think we're gonna cut the video now because that's enough intestines for one day. But Tory, I, I've just got a few questions like quick fire questions that I want to ask you as, as we come to a close. Um, is there anyone who can't take pillow? I'm thinking pregnant women, Children. Well, right now, the, the weird thing is if you're like a child or a cancer patient, you'll have a really delicate G I tract and inserting an endoscope is actually kind of traumatic. And so if we can make it maybe a little smaller, so it would be appropriate for a kid. You might actually be opening up treatment avenues for people that are actually kind of prohibited from them. Ok. And you said that you mentioned this, all that some people call it. The what happens that? Well, we have been swallowing camera pills that don't move but still contain many of the components. We just saw, uh for more than 25 years. And during that time, we've learned how to do it safely and we've also learned that maybe one in five, maybe one in 10 people are going to have trouble swallowing a camera pill. So for us, even if we make it the same exact size and shape, we're going to be missing out on maybe 10, maybe 20% of the population. The thing is though those camera pills because they don't move, they only address about 1% of the population, you can't use them on everyone. And so we're actually talking about going from addressing 1% of people to maybe 80 or 90% of people by making it a virtual experience. And worst case scenario, pill goes rogue within, you know, what, what are the potential health dangers as a result of that happening? Because as incredible as it sounds, we're putting a small drone into the human body and that must come with risks. Well, we had a, we had a pretty amazing uh chat with FDA back in December and they sat us down and the thing that was really thrilling was to hear enthusiasm coming from the biomedical engineers over at FDA, from the regulatory bodies, from the doctors working with them. And they said honestly, this is tech that everyone is kind of hoping for and we want to sit down with your team and figure out how do we do it safely? How do we do it? Right. And what were they concerned about the FDA when you spoke to them? We were concerned about the fact that we use lithium batteries and we were curious if the FDA would be willing to even entertain the conversation. And they actually said, absolutely, sit down. Let's go over your engineering plans. Let's see how you test these batteries. Let's see what happens when a battery fails. Can you contain it? And so far it looks like we're actually on track to being able to turn this into a safe platform. Ok. There's so many more questions I have Pretoria, I'm sure you have Pretoria as well. But thankfully, he will be sticking around for the next two days. So please do feel free to find some at a coffee break, flag him down and continue the conversation because this, as I said is quite an outlandish idea. Thank you for demonstrating it to us today and I look forward to seeing what happens next and thank you so much. Thank you.

Brainstorm AI Singapore 2024

Brainstorm AI Singapore 2024: How AI Is Making The School Bus Run Smarter, Safer, And Greener

Latest video

Ritu NARAYAN, Founder and CEO, Zum Moderator: Matt HEIMER, FORTUNE

Videos (23 videos)

Brainstorm AI Singapore 2024: How AI Is Making The School Bus Run Smarter, Safer, And Greener

Brainstorm AI Singapore 2024: AI in education—fighting cheating and preparing smarter students

Brainstorm AI Singapore 2024: Growth, Wages, Jobs, And Wealth—Debating AI’s Economic Impact

Brainstorm AI Singapore 2024: How AI will impact human creativity

Brainstorm AI Singapore 2024: Saving Lives With AI

Brainstorm AI Singapore 2024: Keeping Up With AI’s Computing Demands

Brainstorm AI Singapore 2024: How to write effective AI prompts

Brainstorm AI Singapore 2024: It Ain't Easy Being Green

Brainstorm AI Singapore 2024: Cutting Data Center Energy Consumption By 50%

Brainstorm AI Singapore 2024: Opportunities In The APAC Market

Brainstorm AI Singapore 2024: Public - Private Partnerships In Ai Governance

Brainstorm AI Singapore 2024: In Search of A Governance Model that works

Brainstorm AI Singapore 2024: Unlocking AI To Boost The Bottom Line

Brainstorm AI Singapore 2024: How AI Is building the cities of the future

Brainstorm AI Singapore 2024: AI will have a $1 trillion impact in Southeast Asia, says Microsoft Asia president

Brainstorm AI Singapore 2024: A Survival Guide to our Superpowered Future

Brainstorm AI Singapore 2024: Singapore tightens AI regulations on political content

Brainstorm AI Singapore 2024: A Global Scorecard

Brainstorm AI Singapore 2024: Strategies for Small And Medium-Sized Economies

Brainstorm AI Singapore 2024: How AI can bridge Asia’s linguistic and cultural divides

Brainstorm AI Singapore 2024: The 1,000-Year-Old Human

Brainstorm AI Singapore 2024: How To Deliver AI-Ready Data

Brainstorm AI Singapore 2024: Swallow This! The Transformative Power Of AI-Enabled Micro-Robotics In the Human Body