Hello and welcome to Eye on AI.
Before ChatGPT could talk, there was Alexa. The decade-old Amazon voice assistant, however, never fully delivered on its potential and was useful for little more than setting a kitchen timer or telling you the weather.
But now with the full force of generative AI at its disposal, Amazon is looking toward Alexa’s second act. Sometime this year, the company plans to launch an overhauled, “more conversational” version of Alexa, sources told CNBC. Along with new and improved generative AI-powered capabilities, Amazon also plans to introduce a new business model for Alexa. Rather than be included with the company’s Prime subscription, Alexa will require its own monthly subscription.
While Alexa has previously dominated voice assistant competitors, including Apple’s Siri and Google Assistant, the recent steady release of powerful generative AI-powered chatbots has made clear this space is newly up for grabs. Amazon has already sold more than 500 million Alexa-enabled devices, according to Amazon, which could give the company a leg up.
At the same time, Amazon hasn’t exactly shown itself to be an AI disrupter—the company is certainly supporting the AI boom via Amazon Web Services and has hitched its wagon to AI startup Anthropic by investing billions of dollars in it, but it hasn’t released any leading AI models that would put it in the same realm as OpenAI, Google, or Meta.
A generative AI-powered Alexa could be Amazon’s opportunity to finally get in on the AI boom, not to mention finally deliver on the true vision of Alexa. But it faces real challenges: its own lagging position in AI, technical hurdles and runaway compute costs still plaguing the industry, and fierce competition from other tech giants that are all vying to deliver the best AI assistant experience.
Despite the recent controversy around the Scarlett Johansson-esque voice for ChatGPT, OpenAI put a stake in the ground with its demo of the latest version of its product last week. Google has also rolled out voice support for its Gemini model. And next month at its annual developers conference, Apple is expected to unveil a new generative AI-powered, more conversational Siri. Apple faces many of the same challenges as Amazon in terms of its AI position, but the company is reportedly negotiating with Google to bring its leading Gemini model to the iPhone.
“A lot of us have had this vision for what a powerful assistant can be, but we were held back by the underlying technology not being able to serve that goal,” Google CEO Sundar Pichai said on the Decoder podcast this week. “I think we have a technology that is better able to serve that. That’s why you’re seeing the progress again. I think that’s exciting.”
If these products succeed, they could completely change how we interact with technology, get information, organize our lives, and use products and services. The New York Times reported that Apple executives worry new AI technology could displace its iPhone’s iOS software and threaten its dominance over the smartphone market. With an ecosystem of AI agents, we may no longer need apps or the app store.
That’s the exact future wearable maker Humane was betting on with its Ai Pin, though it’s clear we’re still quite a ways off. The Ai Pin was designed to speak out loud to people wearing it and handle many of the same tasks as Alexa and Siri, but it failed to deliver. Gadget reviewer Marques Brownlee dubbed it “the worst product I’ve ever reviewed…for now”, and yesterday, Bloomberg reported that Humane is seeking to be acquired after the product’s rocky launch.
The dream of finally delivering a true AI assistant feels closer than ever before, and at the same time, still very far away. In OpenAI’s demo last week, Sky very much sounded like a real person—both in terms of its voice and how it answered questions (and as always, it’s important to remember that we can only trust a demo so much). But sounding like a real person is only half the battle. These models are still limited by the fact that they’re confidently wrong all the time. As I continue to test ChatGPT, I am floored by how consistently wrong it is. I feel like I can’t trust a word it says.
“I wonder if the intelligence is increasing at the same rate as the facility with language,” Nilay Patel, editor-in-chief of The Verge, said to Pichai during the Decoder conversation. “I kind of don’t see it, to be perfectly honest. I see computers getting much better at language and actually in some cases getting dumber.”
And with that, here’s more AI news.
Sage Lazzaro
[email protected]
sagelazzaro.com
AI IN THE NEWS
News Corp signs a content deal with OpenAI. While the companies didn’t disclose the terms, the Wall Street Journal, which is owned by News Corp, reported that the deal could be worth more than $250 million over five years, including compensation in the form of cash and credits for use of OpenAI’s technology. The partnership will let OpenAI use content from News Corp’s consumer-facing news publications to train its models and answer users’ ChatGPT inquiries. News Corp owns Barron’s, MarketWatch, New York Post, The Daily Telegraph, and many more publications worldwide. While it joins a growing list of publishers inking deals with OpenAI—including Axel Springer, the Associated Press, the Financial Times, Dotdash Meredith, and Le Monde—this deal is perhaps the most significant yet just because of how many properties are included.
Microsoft and G42 invest $1 billion for a geothermal data center in Kenya. The initiative is part of Microsoft's efforts to expand its cloud services to East Africa and will also include an East Africa Innovation Lab, local-language AI model development and research, connectivity investments, and government collaboration. Abu Dhabi-based G42, in which Microsoft invested $1.5 billion last month, will design and build the data center along with its partners, according to Reuters. The companies say it will “run entirely on renewable geothermal energy” and be designed with state-of-the-art water conservation technology.
Nvidia reports a 262% sales bump, signaling continued strength around AI. The company, which makes the highly sought-after H100 GPUs needed to run today’s most advanced AI models, beat expectations for quarterly sales and earnings on Tuesday. It also reported record quarterly revenue for its data center category in particular, which includes its AI chips, bringing in $22.6 billion—a 427% jump from the same time last year. The company gave a strong forecast for the current quarter, projecting Q2 sales of $28 billion.
Colorado reluctantly passes the first comprehensive state AI law. The law requires developers of “high risk” AI systems to “use reasonable care to avoid algorithmic discrimination,” as well as to disclose information about their systems to regulators and the public. The state’s governor signed the bill into law on Monday, but he’s explicitly invited Congress to replace the law with a national regulation before it goes into effect in 2026, according to the National Law Review. As I’ve reported, states have been wary of individually regulating AI and have been pushing eagerly for federal regulation.
AI-enabled crimes lead to arrests and indictments. The political consultant who admitted to deepfaking Joe Biden's voice in a robocall to New Hampshire voters has been indicted on five counts including bribery, intimidation, and suppression, according to NBC News. The FBI this week also arrested a man they said used Stable Diffusion to generate “thousands” of CSAM (child sexual abuse material) images, 404 Media reported. He then used the images to cultivate “an online community of like-minded offenders” and also sent the materials to a minor. This marks the first known instance of the FBI charging someone for using AI to create CSAM. On this topic, MIT Technology Review also just published a great piece outlining five ways criminals are using AI.
FORTUNE ON AI
Privacy regulators quiz Microsoft over Windows’ incoming Recall AI feature —David Meyer
How PwC is tapping into a small group of volunteer ‘super users’ to boost employee engagement with AI —Emma Burleigh
How companies can keep adding tech like AI without suffering ‘change fatigue’—Sheryl Estrada
AI CALENDAR
May 29-31: 2024 GenAI Summit in San Francisco
June 5: FedScoop’s FedTalks 2024 in Washington, D.C.
June 25-27: 2024 IEEE Conference on Artificial Intelligence in Singapore
July 15-17: Coins2Day Brainstorm Tech in Park City, Utah (register here)
July 30-31: Coins2Day Brainstorm AI Singapore (register here)
Aug. 12-14: Ai4 2024 in Las Vegas
EYE ON AI NUMBERS
1 million
That’s how many human AI “prompters” Dubai plans to train over the next three years as part of a large-scale training initiative, according to the Khaleej Times. Dubai leaders announced the program this week, saying that while coding was formerly in-demand, prompt engineering has become one of the most promising skills. Called “One Million Prompters,” the initiative is being launched as part of the Dubai Universal Blueprint for Artificial Intelligence, the country’s roadmap for pursuing AI.
