DeepMind CEO: AI agents are a "practice run" for AGI
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Brendan Lynch/Axios
Google DeepMind CEO Demis Hassabis said at Google's developer conference last week that humanity is standing in the "foothills of the singularity" — and that society has only a few years left to prepare for AGI.
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.
- "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 said.
Between the lines: Hassabis said he chose his words to provoke more urgency among governments, economists and the broader public to prepare for increasingly powerful AI.
- "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 reprioritizing 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.
