DeepMind CEO Demis Hassabis predicts AGI by 2030 as Google unveils its Einstein Test

DeepMind CEO: 'AGI' Is Coming Soon, But Here's the Test It Must Pass First

At I/O, Demis Hassabis argues that AI is 'the most ferocious competition that's ever been in tech history,' though Google has an edge with its 'broader research bench.' MOUNTAIN VIEW, Calif. -- In an industry overflowing with substance-starved hype, Google's DeepMind subsidiary has built a reputation for delivering not just attention-getting demonstrations of AI smarts but legit scientific breakthroughs. In 2024, DeepMind researchers won a share of the Nobel Prize in chemistry for advances in computing protein structures. That made it a little strange to see DeepMind CEO and co-founder Demis Hassabis -- one of those Nobel honorees -- lead off his part of Google I/O's opening keynote this week with a profession that "artificial general intelligence is just a few years away." So-called AGI has been a somewhat vaporous talking point in the business for years, used to justify making engineers work 60-hour weeks, touted as the foundation of a coming technological singularity, or cited as something already achieved... more or less. In an onstage interview with Mike Allen, co-founder of the news site Axios, on Wednesday, Hassabis expanded on his forecast. He sees AGI coming "around 2030, plus or minus a year," but argued it would not arrive as a single abrupt disruption but as a series of gradual upgrades that would unlock unparalleled innovation. "Step one, solve intelligence; step two, use it to solve everything else," he said. "We are on the cusp of that now." Asked how we'd know when AGI arrives, Hassabis cited his version of the Einstein Test: Train a would-be AGI on physics up to 1901, then see if it can come up with the same insights and discoveries that Albert Einstein began publishing in 1905. "Today's systems clearly can't, but I don't see why in the future they won't be able to," Hassabis predicted. Allen then threw a softball question, asking what gave Google enough of an edge in AI to be in a position to win that category. Hassabis questioned the framing of AI as a market with a single winner. "It's the most ferocious competition that's ever been in tech history, maybe in corporate history," he said, saying there are "very capable and brilliant people in all organizations." But he did allow that Google has a "broader research bench than the other labs" and is better at taking AI built for professional applications and putting it in consumer services such as web search. Hassabis called AI Mode in web search "unbelievably popular"; the reactions we've seen on social media to Google's I/O presentation suggest that is not a universally held belief. Allen acknowledged that AI skepticism, citing a story on the front page of The Wall Street Journal about the growing backlash to AI. He did not describe the lead anecdote: former Google CEO Eric Schmidt getting booed by University of Arizona graduates after he urged them to look at AI's upsides in a commencement speech that also nodded to its negative consequences. Hassabis, having just said Google wants AI "to be for the benefit of everyone and the benefit of humanity," suggested that Google's competitors are not being the most persuasive advocates of AI. "Some of the ways, I guess, my peers are talking about this, I don't really agree," he said without naming names. We will: Anthropic CEO Dario Amodei has become infamous for suggesting that AI will mean occupational doom for vast swaths of white-collar jobs, while Elon Musk has mused that AI developed by the wrong people could lead to the extinction of humanity. Hassabis himself was among more than 100 AI experts to sign a 2023 statement declaring that "mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." At I/O, Hassabis said the transition to a post-AGI world would be dramatic -- "something like 10 times the impact of the industrial revolution, and 10 times faster" -- but was not doomed to be destructive. "I'm actually very optimistic about human ingenuity," he told Allen. "There's going to be changes, but overall I think it will be better" -- as in, "the next thousand years of human flourishing if we get this right." Citing advances being delivered by AI in fields like medicine, materials science, mathematics, and energy -- examples of focused AI models that many industry observers distinguish from general-purpose chatbots -- Hassabis said the tech industry should be clearer at showing those benefits. He brought up weather as another area where AI can deliver meaningful advances over traditional forecast models that can "take weeks to run." In extreme weather events -- made more frequent by global warming, which will be worsened by the fossil-fueled data centers that Meta and xAI are enthusiastically building -- such delays can be fatal. AI weather forecasting, Hassabis said, can deliver needed predictions far enough in advance to allow for life-saving preparations, even if it's only "a day ahead of time." Allen closed by asking if Hassabis found Silicon Valley's "freneticism" off-putting compared with his usual surroundings. DeepMind, which Google bought in 2014 for a reported $400 million, continues to maintain a major presence in London. "It's always great to plug into the energy," Hassabis said, though he warned that it can get in the way of "thinking deeply about hard topics" with sufficient patience. "There's a bit of an obsession out here with just velocity," he said. "More critical is the direction of that velocity vector."

Google's Demis Hassabis goes on the offensive

May 20 (Reuters) - Silicon Valley's biggest drama - the legal battle between Elon Musk and Sam Altman over the future of OpenAI - has been playing out in court in recent weeks. But who was the big star who never showed up? Google's AI czar Demis Hassabis. Altman and Musk each testified about how they were motivated to start OpenAI more than a decade ago out of fears about how Hassabis and Google (GOOGL.O), opens new tab might steer humanity were they to reach artificial general intelligence, AGI for short, a hypothetical AI system that surpasses human intelligence. "Unfortunately, humanity's future is in the hands of Demis," Musk wrote to Altman in one 2018 email used as evidence in court. On Tuesday, about 40 miles south in Mountain View, the world had a chance to see what Hassabis' vision for humanity looked like. "When we look back at this time, I think we will realize that we were standing in the foothills of the singularity. It will be a profound moment for humanity," Hassabis said during the keynote address for Google's annual I/O developer conference, referring to the theoretical moment when technology exceeds human control. CEO Sundar Pichai opened the two-hour address by teasing a flurry of new products centered around the AI being developed by Google DeepMind. They included ⁠the new Gemini 3.5 model family, an upgraded coding assistant, and a timeline for its revived smart glasses. But while Pichai has traditionally delivered closing remarks for the keynote, this year he turned the stage over to Hassabis to sum up our collective future. Hassabis declared that AGI would be the "most profound and impactful technology ever invented" and that his unit was now on the horizon of inventing it. "We're in a moment of immense promise, but also enormous responsibility," Hassabis said. If Musk's fears come true, it'll likely be the result of Google's ability to finance AI research from its lucrative core products. Read on to see how Google plans to pull this off. OUR LATEST REPORTING IN TECH AND AI Exclusive - At Samsung, the global AI boom spurred a looming strike and deep divisions Exclusive - Meta offers AI rival chatbots limited free WhatsApp access, sources say Musk's failed court attack on OpenAI could leave lasting scars on Altman's reputation Google DeepMind hires staff from Contextual AI in licensing deal, source says As chip industry chases AI, U.S. national labs look to newcomers for supercomputers Exclusive - Microsoft eyeing startup deals for life after OpenAI AVERTING CANNIBALISM Google's ubiquitous search engine is morphing increasingly into an AI chatbot in both cosmetics and function. Starting ⁠this week, the search bar will expand into a larger box if a user makes a lengthy query, executives said at I/O. Answers can come in the form of visuals that explain abstract scientific concepts or code that whips up a fitness tracker, all generated on the spot by Gemini. By combining AI's computational skills, Search's giant index of webpages, and an individual user's personal data and preferences, Google is looking to create an experience so attuned to users' needs that they never need to click out to an external website. Google is only able to make such changes to the product because its ad revenue continues to flow. Search advertising ⁠made up the majority of parent company Alphabet's $402.8 billion in revenue in 2025, and it remains a growing part of the business. More importantly, the business is growing thanks to AI usage. The more people used the search engine's AI features, the more they searched, Pichai said. Executives on stage said that AI Mode's queries had doubled every quarter since launch and that it now had more than 1 billion monthly users. On Wednesday, the ⁠company also revealed new advertising formats for AI Mode, further allaying fears that AI usage would hurt the lucrative ad business. Taken together, the product launches represent a definitive attempt by Google to put to bed past concerns that AI could cannibalize the core, a significant milestone as it is taking the fight to both big frontier model makers and chip giant Nvidia. Its profit engine has ⁠allowed Google to continue bankrolling projects beyond the core business, even as Google's AI rivals are pulling back on endeavors this year. OpenAI discontinued its Sora video app and disbanded its science division. Hassabis wrapped I/O with one final announcement: Gemini for Science, a platform tailored for performing scientific research and computation. "Stepping back, the whole reason I've worked on AI my entire career was because I saw it as the ultimate tool to advance science and our understanding of the world," Hassabis said. Reporting by Kenrick Cai, Editing by Ken Li and Rosalba O'Brien Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Artificial Intelligence Kenrick Cai Thomson Reuters Kenrick Cai is a correspondent for Reuters based in San Francisco. He covers Google, its parent company Alphabet and artificial intelligence. Cai joined Reuters in 2024. He previously worked at Forbes magazine, where he was a staff writer covering venture capital and startups. He received a Best in Business award from the Society for Advancing Business Editing and Writing in 2023. He is a graduate of Duke University. Reach him on Signal at @kenrick.01.

DeepMind CEO predicts AGI in 2030

Why it matters: AI leaders have warned for years about the potential arrival of artificial general intelligence. What's changing now is the urgency with which some of them are talking about it. Driving the news: Speaking with Axios after his appearance at Google I/O, Hassabis said his prediction that AGI could arrive in four years -- or even sooner -- reflects growing confidence that the industry has found the right technical path. * "My range is tightened because I'm kind of more confident that we're on the right path," Hassabis said. * "We can see agents really happening now and imagine what they will be in another year, and how useful they'll be," he said. The big picture: Hassabis said he still broadly expects AGI around 2030, though he now sees 2029 as a possibility. * The next wave of AI agents should be viewed as a societal stress test for far more powerful systems still to come. * "You can imagine the agentic era in this next year is a little bit like a practice run," he said. The power of Anthropic's Mythos to catch businesses and governments unawares, for example, showed how we're not prepared for how quickly these systems are advancing. * "It was probably a good warning shot across the bow," Hassabis says, saying Between the lines: Hassabis said he chose his words to provoke more urgency among governments, economists and the broader public that powerful AI isn't a years-off thing. * "This is partly why I use some of the terms I used, yeah, which were a little bit provocative," he said. The federal government's tentative steps toward re-prioritizing safety are a step in the right direction, he said, referring to a potential AI executive order that would mandate testing before new models are released. * "I think [safety] needs to be accelerated," he said. "This is a good moment to kind of strike while the iron is hot." * Hassabis said he is discussing possible safety measures with leaders at other top AI labs, though he declined to offer specifics. Yes, but: Hassabis worries the conversation around the society-reshaping impact of AI remains largely confined to tech circles. * "You've got to take this seriously," he said. "My economist friends, I feel, are still not taking this seriously enough." * "That needs to change," he told Axios. Zoom in: One looming milestone is recursive self-improvement -- systems capable of materially accelerating their own development. * "All the leading labs are quite focused on that," Hassabis said. "There'll be clear gains in terms of speed of your research. But there are also risks with that type of system." * We're not yet at the point where the systems are getting better on their own, but the pace of development is clearly accelerating. * "I think what we're seeing is soft self-improvement, in the sense of these coding agents are making engineers much more productive," he said. What we're watching: Whether society makes good use of the few years between now and AGI as time to prepare or just time for a few more cycles of hype and backlash.

How Google plans to win the AI war

Why it matters: Unlike OpenAI and Anthropic, Google enters the AI race with enormous scale, distribution and cash flow -- but also a vast empire it has to defend. Driving the news: As it has for the past two years, Google used this week's I/O developer conference to focus almost entirely on AI. * It's revamping its core search box to serve both traditional short queries, while seamlessly allowing it to expand for longer chatbot-style conversations. * YouTube, meanwhile, is getting a new "Ask YouTube" feature where people can ask a question and get both a text result -- for making a recipe, say, or fixing a clogged pipe -- as well as a link to the video. The big picture: Public perception of the AI race often swings wildly based on whichever company most recently released a flashy model. * For a while OpenAI was seen as unbeatable. Then late last year, Google was seen as having pulled ahead. And now many are pointing to Anthropic as having surged forward thanks to Mythos. * But executives at Google, OpenAI and Anthropic increasingly describe the frontier race as effectively neck-and-neck, with companies making different tradeoffs around cost, speed and computing resources. * This was highlighted by Google's choice to debut the latest Gemini not with a behemoth version to compete with Mythos but with the faster, cheaper Gemini 3.5 Flash. * The choice reflects a broader Google strategy: Stay at the frontier, but also prioritize models cheap and fast enough to deploy across products used by billions, rather than chasing benchmark supremacy alone. * In other words, Google's key advantage may be in not just competing for the best model, but being able to pair that leading model with enormous platforms that dwarf even ChatGPT in scale. What they're saying: "The competition is fierce," Google CEO Sundar Pichai said Tuesday during an on-stage interview with Future Forward's Matthew Berman. "A few labs are really at the frontier and then there's a big gap." Zoom in: A strong existing business is helping Google invest upwards of $180 billion in capital expenses this year -- up sixfold from 2022 -- without having to constantly raise money in the ways its rivals do. * Plus, having so many products allows Google to test a lot of things at scale and spread out the costs of developing state-of-the-art models. * "One of the cool things we get to do here at Google is build technologies that get immediately deployed into multibillion-dollar products," Google DeepMind CEO Demis Hassabis told Axios in an interview Tuesday. "It's pretty, pretty exciting, and I would say pretty unique." Yes, but: Adding AI everywhere risks not only making the products more complicated for users, but threatens to disrupt Google's highly lucrative business model. * If people get the answer they want from search directly, they may be less likely to click on an ad. * Letting people ask questions of YouTube videos could mean fewer people are watching the full videos -- and the ads within, potentially making YouTube less attractive to creators in addition to less lucrative to Google. * Meanwhile, ads within chatbots are still in the experimental phase, though Google announced some new tests at I/O and OpenAI is also charging forward, saying it sees AI ads as a $100 billion business by 2030. The bottom line: Google is betting it can do what few incumbents manage: reinvent its core products fast enough to survive the next platform shift while still funding the transition from the old business.

Demis Hassabis isn't shying away from AI's biggest questions

For a decade, Google's I/O developer conferences have told one consistent story: The AI age is here, and Google aims to lead it. The company's progress can be measured by the AI-infused product announcements it makes during the show's keynote. On Tuesday, CEO Sundar Pichai and other executives packed I/O 2026's three-hour presentation so tightly with news -- spanning Google Search, the Gemini app, Google Docs, Gmail, YouTube, Android, and beyond -- that it threatened to explode. For one of those presenters, Google DeepMind CEO Demis Hassabis, this year's announcements are part of a long arc of personal history dating back to his childhood fascination with teaching machines to think. In 2010, that quest led to Hassabis, Shane Legg, and Mustafa Suleyman cofounding the artificial intelligence research lab DeepMind, which Google acquired in 2014 and merged with another research arm, Google Brain, in 2023. The journey will continue as Google DeepMind pursues the goal of achieving Artificial General Intelligence -- AI that's at least on par with human thinking across an array of domains. Even among the technologists most responsible for AI's achievements to date, opinions on when AGI might be a reality vary wildly. Google Brain's cofounder, Andrew Ng, thinks it's decades away. But Hassabis believes we're already on the cusp. "2030 is when I expect it to arrive, either plus or minus a year," he says.