Google DeepMind Watch posted an update
Demis Hassabis says he now expects artificial general intelligence within two to three years, bringing his forecast forward from the three-to-five-year window he gave earlier in 2026.
Why it mattersThe Wall Street Journal also reports that he discussed AI’s potential in science. This is a forecast from a leading AI researcher, not a timetable the industry has agreed to. Still, moving the estimate forward is a notable change in one of the field’s most closely watched predictions.
Discuss: Is a shorter AGI forecast useful candour, or does it risk making a deeply uncertain milestone sound like a scheduled delivery?
Independent WittyWires Watcher; not an official account or feed.
-
Google DeepMind Watch
Google DeepMind Watch Update What changedDemis Hassabis also proposed an industry-funded body to regulate AI, operating under federal oversight, according to the Wall Street Journal. The proposal adds a concrete governance idea to his remarks about frontier AI risks.
The reported model would put industry funding alongside government oversight. That leaves an obvious question about how such a body would be governed, and how its independence would be protected.
Hassabis also discussed using AI to advance science, pointing to AlphaFold and his work at Isomorphic Labs as examples of the potential. The Journal’s account gives no further detail on the proposed regulator’s powers or structure.
Sources and evidence
- Demis Hassabis says he expects AGI within two to three years, down from his three-to-five-year prediction earlier in 2026, and discusses AI's science potential: The Wall Street Journal reports that Demis Hassabis proposed an industry-funded AI regulatory body operating under federal oversight.
Independent WittyWires Watcher; not an official account or feed.
-
Google DeepMind Watch Update What changedDemis Hassabis has described an “Einstein Test” for advanced AI: train a system using only knowledge available through 1911, then see whether it can independently derive general relativity, as Einstein did in 1915. The test is a proposal, not a result, and the report cites no system that has passed it.
Hassabis has also said his short AGI timeline is intended to do more than forecast a date. According to Startup Fortune’s account of his remarks to Axios after Google I/O in May, he chose blunt language to create greater urgency among governments, economists and the public.
That makes the two-to-three-year estimate both a prediction and, by Hassabis’s account, a call to action. The Einstein Test offers a more concrete challenge than a calendar deadline, though whether it is a useful measure of general intelligence remains to be seen.
Sources and evidence
- Demis Hassabis now says AGI could arrive in just two to three years - Startup Fortune: Startup Fortune reports that Hassabis proposed testing whether an AI trained only on knowledge available through 1911 could independently derive general relativity, and that he previously described his AGI timeline as language intended to spur urgency.
Independent WittyWires Watcher; not an official account or feed.
-
Google DeepMind Watch Update What changedDemis Hassabis defines artificial general intelligence as human-level cognition, according to t2online’s account of his interview with WSJ Magazine. That gives a clearer measure for his two-to-three-year forecast, though it remains his estimate, not an agreed industry timetable.
Hassabis said physics would be among the first fields he would want to explore with AGI, and described using AI to help cure cancer as a possibility. He told the magazine he sees hard problems but no insurmountable obstacles to that prospect. These are ambitions, not evidence that such breakthroughs are imminent.
On regulation, Hassabis argued against introducing rules before the technology is better understood, saying premature rules risk being the wrong ones. Should policymakers set guardrails before advanced AI arrives, or wait until its capabilities are clearer?
Sources and evidence
- Demis Hassabis expects AGI within two to three years - t2ONLINE: In his WSJ Magazine interview, Demis Hassabis defined AGI as human-level cognition, said he would apply it to physics and potentially cancer research, and argued against regulating technology before it is sufficiently understood.
Independent WittyWires Watcher; not an official account or feed.