Name: Brainstorm Tech 2023: Unlocking A.I.’S Potential
Uploaded: 2023-07-10
Duration: 19 min 53 s
Description: Jonas Andrulis, Founder and CEO, Aleph Alpha Antonio Neri, President and CEO, Hewlett Packard Enterprise In conversation with: Jo Ling Kent, Senior Business and Technology Correspondent, CBS News and Co-chair, Fortune Brainstorm Tech

Jonas Andrulis, Founder and CEO, Aleph Alpha Antonio Neri, President and CEO, Hewlett Packard Enterprise In conversation with: Jo Ling Kent, Senior Business and Technology Correspondent, CBS News and Co-chair, Fortune Brainstorm Tech

Brainstorm Tech 2023: Unlocking A.I.’S Potential 

Conferences

Jonas Andrulis, Founder and CEO, Aleph Alpha
Antonio Neri, President and CEO, Hewlett Packard Enterprise
Jo Ling Kent, Fortune Brainstorm Tech 


Jonas Andrulis, Founder and CEO, Aleph Alpha
Antonio Neri, President and CEO, Hewlett Packard Enterprise
In conversation with: Jo Ling Kent, Journalist and Co-chair, Fortune Brainstorm Tech 


Our next conversation will be our first on one of the hottest topics in tech artificial intelligence. And we're excited to dive right in. Let's start with the big picture ai trends and more specifically how they could apply to enterprise as opposed to end user solutions and help create strategies that scale. We'll be interviewing two industry veterans who are working in tandem to maturing technologies to the test. Joni Andreis is the founder and CEO of Ali Alpha. Joni's career has been entirely dedicated to R and D of AI at of Alpha and its focus is on channeling multimodal input to help businesses control their own data knowledge and create value. Antonio near is the president and CEO of Hewlett Packard Enterprise. Antonio has been with HPE for over 25 years and led it through breakthrough innovations and acquisitions, particularly in areas such as hybrid it, edge computing and intelligent data management. Joni and Antonio are joined by our brainstorm tech co chair, Jolene Kent. Good afternoon. Thank you Jessica. It's great to be here. How's everyone feeling good, good, good. I did the Bob club this morning thought I was going to die but feeling good now. Glad to be almost no jet like Yes, exactly welcome. So I want to just start with the basics and kind of set the table here A I is obviously fully accessible to all of us consumers now, right? So how would you describe where A I is right now? And where is the opportunity in 30 seconds or less A I is phenomenal for doing your homework or create copy for like marketing or like images that sound kind of cool, but it's really difficult to use it for things that are critical or that rely on compliance issues. So I think we are, we are in a playful area where we can play around with that, but it's really difficult to get it right where we have a lot of responsibility, Antonio, do you agree? I do agree. In fact, we are, you know, the analogy I want to make is that I play a lot of soccer. I still play in my fifties and we are in the first minute of the game, we are in the first minute of the game when it comes down to applicability in the enterprise with clear use cases where we can actually deliver value from the data. But no question A I will be the most disruptive technology in our lifetime. Where do you see the opportunity in many places? I think, you know, A I has the potential to really solve some of the biggest societal challenges we all live, Vice President Gore, talk eloquently about the challenge of our climate. So when it comes to, I, I think there are three fundamental things we have to think about it. Number one is obviously use it responsibly and there is a lot of chatter about that and what that means. Number two is data privacy. And obviously in the enterprise space, there is a lot of concern about how you're going to train these models and how you're going to deploy this model, maintaining the, the privacy about data. And last but not least is the environmental impact AI is going to have which we're going to get to later. I really want to dive into that because it's something I hadn't really thought about much in my reporting on A I just a friendly reminder. I want to go to audience questions. So if you, if you all have questions, you know, take note and then we'll call on you throughout the session and we'll need you to stand up and identify yourself, please. So, so we're, we're here in part because you're collaborating on HPE Green Lake. It's for large language models. And you basically want to enable companies to deploy large scale A I with super computing cloud capabilities. That is a mouthful. What does it actually mean? And why do you think that partnership is going to move the ball forward? Antonio? Yeah, maybe we should start a Joe by explaining what HP Green Lake is HP Green Lake for Hewlett Packard Enterprise is a cloud platform where we offer intelligent edge services, hybrid cloud services and now A I services. So we join us. We just announced two weeks ago, the first enterprise available large language model available in our cloud in the second half of this year. So the customers can come and train privately their data and start training and tuning their models that eventually can deploy in the inference, which is could be any place in their enterprise. It can be in a manufacturing floor, it can be in a hospital, it can be in their own data center for that matter. So for us, it was a pivotal moment because HP is today the largest vendor when it comes down to supercomputing power, we today have deployed last year, the largest supercomputer ever built in humankind. It's an ex scale system that has the ability to process a quintillion operations per second and to put it in context, that's 40,000 gps and 21,000 cps which you know, has a lot of challenges in the way you manage it. But we are now in the process to deploy even a larger system which has 60,000 gps and will be twice the performance of that system. So the question is, how do we make it available to everyone in a way they can train their data and pay as you go? And that's the goal with that announcement with Iness. Yeah. And we're focusing that. So what, what our goal is and why I like this partnership is because we share the same goals. We want to bring sovereignty to the enterprise and two governments. Exactly. So we want to empower them. We want to give them the ability to build trustful systems in regulate highly regulated environments for governments, for security, for health, for legal, for finance. And for that, we need a partner that shares our interests. And I think that's why I'm super happy about this because precisely what you said is the challenge. Where is my data going? How can I make sure that I can leverage information that should never leave a certain restricted zone, but I still can make sure that I can leverage all the power of Janii. And this is really a tremendous opportunity that we have here. You, you said trustful, I mean, trust is one of the biggest, most difficult issues to tackle in tech in general, but especially in A I, right? And you are both really going for it to try to make it work for your, for your customers. So it's obviously clear we've seen it already the downside of AI and we've seen the dangerous impact that it can have. So how do you guard against it? How do you be the trustful provider of this technology and mitigate the downside impacts? So, first of all, it's just tech, right? So we should not over interpret the abilities and what we're seeing there. Responsibility. I am convinced can only be taken over by humans. It doesn't matter how great A I will become, it cannot take over responsibility. So our design philosophy is not, how can we build a chat bot that gives you the right answer. But how can we give superpowers to the expert? How can we help the experts? How can we help all the great people that are in enterprises and governments to take responsibility and have like two s five s productivity? And I think this is our design philosophy and this is very different from a lot of like see products that basically blurb out an answer and then you basically have no alternative to trust it, no idea where it came from. Yeah, you have no idea and you basically have to, if you want to verify it, you have to open a second window and start a search. So how do you guard against that? Well, how does the product work to make it trustworthy? So like Antonio said, first of all, we can leverage proprietary data, a secure environment. That's where it starts. But then based on our research, we can also link this data to its sources. So we can and you probably have seen this on like references. So this is a link to an URL. The interesting thing is that there's no guarantee that this URL actually backs your claim. And in many instances, it doesn't. So if you click on that URL, it's like a long document you can read through all of that. And in the end, you'll find nothing in this document really supports that claim. This is how a lot of systems work currently because they are adding these links based on embeddings and embeddings have no guarantee what we are doing is we are linking to the precise situation, the precise words or sentences that actually back a claim and not only in the positive but also in the negative in the sense that if there's something that's conflicting with a certain claim, you can also see that can you, I can show you yes, I brought some examples at sea. Here we go. Is that visible? All right. Awesome. So here is Heidelberg, this is our our headquarters. So if you're ever in Germany, you visit and it's multi mode like we've seen this and you see the prompt is this image and the task to describe the image. And then the answer is a beautiful sunset over a city. So let's look at how this is proven, how we can prove that. And now you can see sunset is proven by this part of the image and city by this part of the image. And you can also see some kind of blue parts picking up in the image, right. So on the city part, the castle on the very left left is colored in blue because it's not a city, it's a castle, right? So now I can see what is conflicting, what is supporting my claims and I can do this based on all my data sets, all my proprietary data. And we're going even one step further in that we are giving this A I the opportunity to talk to you because what you actually need is let's say you've got 10 million documents. You don't need links to 10 million documents with parts that are supportive or conflicting. What you need is a system that tells you something like this. So here, I'm asking what's the former capital of Germany? And then the reply is they are contradicting answers saying the former capital of Germany is Bonn or Berlin. And for both of those, I got claims I got sources. And so this is what we showed together with HP at HAFA me was a great success on as a use case in manufacturing, actually based on secure and secure environment on proprietary highly complex technical information and helping the expert working with the system, working with the robots that given comments that was precisely execute against their preparative information. How important is it to collaborate in these kinds of partnerships? Antonio to bring to market the A I that we already know we want and need whether you're a corporation, organization or an individual? I mean, is, is it basically necessary at this point for a company like HPE absolutely. And if you think about the use case, the amount of computational power that took to get to that precise model is, is astronomical, you will not think about it. So the stage is that when you think about today, the public cloud, as we know it, normally there, you put a lot of workloads, a lot of applications and a lot of customers that they are trying to use every possible resource. But when you go to a model like this, it's actually one word load, one A I model using an enormous amount of accelerated computing and optimizing the model with a compute is absolutely critical in order to deliver the performance and the accuracy. And then ultimately, the completion of the model because the most important metric is when he starts the model, he has to finish the model and now stop because otherwise restarting causes a lot of inaccuracies and more power consumption and waste. So that's why the partnership is important because we both can collaborate and deliver a better experience. You mentioned waste. Yeah, you mentioned sustainability earlier. I mean, how much given all of the power that's required to make these models work and this technology actually flow right and work, right? We're talking about a lot of energy. So Antonio, I know you're really passionate about this, but how do we guard against, you know, kind of working backwards when it comes to energy, the environment, climate change, you have to start from the beginning, right. So if you think about some of these computational power, it's just a simple rack system that has a few 100 GPS can consume a half a megawatt one multiply. In the case of the Oakridge National Laboratory have 75 rack systems to deliver the 1.1 flop that's 35 megawatts just for the compute side. Then you need all the power to do the cooling and the storage and so forth. So you have to design this from the beginning and then ultimately think about the circular aspect of this. So our A I cloud that we are building actually is going to be 99% renewable energy. But the waste is going to go to actually power greenhouses. And the waste itself is what we talked about hot water, right, hot water and obviously energy diffusion. But, you know, we have done projects with the National Renewable Energy Lab. Actually, they take the hot water and they actually provide heat into the building in Colorado without wasting more energy and then ultimately melt the snow in the parking lot. But ultimately, we can actually define a carbon negative negative carbon and that requires a whole engineering ecosystem up front, not after the fact and that's what HB is great about it. It's, it's that part to me is extremely fascinating. In addition to the trust that we need to have in our products, I'd like to open it up to the audience. Are there any questions so far? Anyone? Oh, ok, great. I see it in the back. Please please identify yourself. My name is Raj Mohan. I'm the Chief Digital Officer at Ascension, a large health care system. I have a two bar question. Um, health care is an area that where the experts are burnt out. They're quitting. And I'd love to your perspectives on the role of A I and A I second part. My daughter's a senior in writings, senior majoring in data science minor and com science freaking out slightly about chat GPT. What advice would you give her? Oh, let's start with the second one first. Maybe he can go first. So I think if you're just starting your career, it's a good advice maybe to become a plumber because robotics is lagging behind a lot. Um I think for knowledge work, this will change a lot. Knowledge work over the next few years will change a lot. So don't focus on memorizing knowledge, focus on navigating innovation and, and helping helping humans, helping make world a better place, right? So maybe more on like what, how to live a great life, how to, how to drive innovation, how to juggle change. Are these kinds of skills that are not classically taught in universities that much will become much more important? I will say, obviously, our education system is going to change profoundly, but you know, I was just still doing things by hand, so you can learn the basics. I mean, I grew up in Argentina, I went to middle school and high school where I was working full time and then to college and I had to make photocopies of the books in order to get access to the textbooks. And, and I tell you that was the best learning experience because you put the effort. Now, obviously, I will change how we learn going forward. But the fundamentals cannot be replaced in making you a better engineer or whatever a career you take forward. Now on health care, I'm actually a director in the second largest insurance company in the United States. It's called Evan Health. In the past, it used to be called anthem. So for those of you who have an insurance and we are basically every board meeting, we're talking about A I. But in the context of reducing costs, health care today has an enormous waste in terms of the way we manage claims and the way we provide health care to all citizens. But obviously focus on preventative health, you know, preventative medicine and some of the projects we are doing ourselves at HPE, particularly in Germany where I is coming from. We have deployed A I what I call a swamp learning technique where every year the entity takes the scan of 30,000 citizens. So that by 2026 2027 we can actually model the entire brain. So we can in the protein about Alzheimer's and dementia. But because of this data privacy, we actually connect the swamp learning so that all the researchers can get access to the insights, not the data. So people can share real time insights without sharing the data. And that's also another way how A I is going to be deployed connecting super important is important, how we share insights, not the data. And this is I think is going to change the way we work together. I'm going to get to our one last question here. Please stand up and identify yourself. What's your question, Jeff Roberts Fortune. My question is, you know, so there's a bit of hype around AI might be a lie. But when you said this 99% carbon neutral or green or carbon negative, I mean, that frankly seems a bit too good to be true. Is that just aspirational? And there's just some way to audit this and how do we know for sure. That's true. It is 100% hydropower, it's 99% hydropower plus some, you know, generation backup and that water gets recirculated. And then ultimately, we take the hot water and we power other secondary sources. But today, the data center is being built in Quebec and is 99% energy through hydropower. And it's even maybe one step further. I think you mentioned that is not only are we being incredibly careful with the resources, we are spending now. But this technology will help us tackle the challenges of the future. I don't think we can solve climate change without innovation. That's right. And one of the major drivers for innovation will be A I. Yeah. And that's why A I will unlock a lot of the research and the knowledge needed to solve some of these societal challenges. If you look at the system, at the Knoxville laboratory, a lot of the research there is about climate is about life science. If you think about the vaccine on COVID, we used to take many hours to crunch, we actually did it in eight minutes and that allowed them to do 1.2 million more scenarios to get to the spike protein. So clearly, when applied the right way and responsibly can have the potential to solve some of the biggest challenges we ever faced. Well, Antonio, thank you both so much. I think we could talk about this for another hour, but we've got to end here. I really appreciate it. Thank you. Thank you.