|
||||
|
||||
|
By Aaron Mak |
||||
|
||||
Demis Hassabis | Google DeepMind handout. |
||||
Demis Hassabis is the co-founder and CEO of Google DeepMind, an AI research laboratory responsible for development of the Gemini large language model and AlphaFold, a system that revolutionized the process of analyzing protein structures for drug discovery. Hassabis and DeepMind’s John Jumper won the 2024 Nobel Prize in Chemistry for their work on AlphaFold. The invention of AlphaFold led Hassabis to establish a spinoff of DeepMind known as Isomorphic Labs, an AI pharmaceutical company. Hassabis talks to us about his goal of eradicating all diseases, and why he’s blown away by models that can understand the physical world. This interview has been condensed and edited for clarity. What’s one underrated big idea? The potential for AI to be the ultimate tool for advancing science and understanding the universe around us. I’ve always believed that AI will be the most transformative technology in human history. This is now widely recognized, but the profound impact it will have on the pace of scientific discovery — and the impact that will have — is still underrated. It will be at least 10 times [as impactful as] the Industrial Revolution. I’ve spent my entire career building AI because I think it will help us solve some of the biggest problems society faces, whether by accelerating drug development or finding new sources of clean energy. We could reach a point where, for the first time, resources might no longer be the limiting factor for human progress. It could be an amazing new post-scarcity era for humanity. We’re already seeing incredible progress at Isomorphic Labs using foundational AI models to enable “science at digital speed.” The goal there is to reduce the time it takes to design a drug from years to months or even weeks, and ultimately, one day, solve all diseases. What’s a technology you think is overhyped? There is a certain amount of overhype around AI in the near term from some quarters, but in the medium-to-long term, the enormous impact it is going to have is actually still underappreciated. Today’s systems have some impressive capabilities — they can create videos, explain concepts across a vast range of topics and converse in different languages. But they can still sometimes fail at elementary math problems and other trivial tasks. We see no fundamental scientific barrier to AI eventually becoming AGI, a truly general-purpose system that can do all of these things consistently. I suspect it could take a few major breakthroughs as well as further scaling of existing systems to bridge the gaps on capabilities like long-term memory, continual learning and true creativity. |
||||
|
New from POLITICO
Tracking the forces shaping politics, policy and power worldwide, POLITICO Forecast connects developments across regions and sectors — including key global moments and convenings — drawing on POLITICO’s global reporting to help readers see what’s coming next.
|
||||
What could the government be doing regarding technology that it isn’t? AI will impact the whole world, and in order to fully realise its benefits and mitigate any potential risks, we need deep, sustained dialogue between governments, companies and civil society. International forums like the AI Summit series — first started in the U.K., with a fourth taking place in India next month — are good opportunities for maintaining momentum on this effort. AI is a borderless technology, and we should try to foster international cooperation on its most serious risks. I see two main priorities here — firstly, adopting technical and governance standards internationally as this will enable the responsible diffusion of AI to benefit society and secondly, advancing the collective understanding of AI risks to help inform policy decision-making around the world and ensure a consistent approach. What book most shaped your conception of the future? The Culture series by Iain Banks. I think it is the best depiction of a post-AGI world, where AI systems coexist with humans and alien societies, and humanity has spread consciousness across the galaxy and sort of reached maximum human flourishing. It paints a picture of a civilization that has solved resource constraints and can focus on art, philosophy and exploration. That brings bigger philosophical questions about the meaning and purpose of life, but it is an amazing, compelling future that I would hope for humanity. What has surprised you the most this year? We’re seeing remarkable progress with “world models,” or models capable of understanding how the physical world works. Just by watching millions of videos, our AI models, like Genie 3, seem to be intuiting some notion of physics, such as how liquids flow or objects cast shadows. Accurate world models will be essential for AGI systems to plan, reason and act in the real world, whether in robotics or as assistants on our smart devices. Having designed AI and graphics for video games in my early career, I remember how painstakingly hard it was to program these mechanics by hand. Seeing these models reverse-engineer physics just by observing is mindblowing. |
||||
|
||||
|
||||
A new California bill would prevent public officials, lobbyists and political candidates from using insider information to place bets on platforms, like Polymarket and Kalshi, which let people make predictions on real world events, POLITICO’s California Decoded team reports. Prediction markets often allow users to bet on political developments, like whether there will be a government shutdown or what a public official will say in an upcoming speech. The issue of insider trading also led House Democrats to introduce a similar bill on the federal level, due to suspicions around bets that were being placed on prediction markets shortly before the U.S. arrested Venezuelan leader Nicolás Maduro. Republican Assemblymember Kate Sanchez, who introduced the California bill, says that it will close a “loophole” in state law, since California does not explicitly prohibit insider trading in prediction markets. “The law needs to be clear: public service should never be used for personal profit,” she told Decoded. |
||||
|
||||
|
||||
Russian networks have been spreading false narratives surrounding the Justice Department’s recent release of files involving Jeffrey Epstein, POLITICO’s Morning Cybersecurity reports. Research from the Institute for Strategic Dialogue discovered that the Kremlin-backed group Storm-1516 has been producing fake news articles that falsely link French President Emmanuel Macron to Epstein. Some contained screenshots of forged email exchanges between the two. Influencers on X helped to amplify the reach of the articles. Online disinformation about Macron and Epstein has become so prevalent that it’s led French government officials to blame Russian-linked groups for fueling the unfounded rumors. |
||||
|
||||
|
||||
Screenshot/X |
||||
|
POLITICO Pro
POLITICO Pro Briefings give subscribers direct access to in-depth conversations on the policy issues shaping government. Led by POLITICO reporters, these live interactive sessions go beyond the headlines to explain what’s happening, why it matters, and what’s coming next.
|
||||
|
||||
Stay in touch with the whole team: Aaron Mak ([email protected]); Nate Robson ([email protected]); Ruth Reader ([email protected]); and John Hendel ([email protected]). |
||||
|
Follow us on X
|
||||
|
Follow us
|
||||
