Google DeepMind co-founder and Nobel Chemistry laureate Demis Hassabis said in an interview published by WSJ Magazine on October 21 that he now expects artificial general intelligence (AGI) to arrive within just 2–3 years, around 2028–2029.
- Three jumps in one year: At Davos in January, Hassabis still put AGI 5–10 years away; earlier this year the estimate was 3–5 years; now it has shortened to 2–3 years. In 18 months, the timeline has shrunk by more than half. His explanation was brief: "It's soon. I'm updating nearer and nearer."
- A very high bar: Hassabis defines AGI as a system with the full range of human cognitive abilities, not high scores on individual tests. His "Einstein test" asks: trained only on knowledge available before 1901, could an AI independently derive Einstein's 1905 theory of relativity? It tests scientific creativity — posing new questions and hypotheses — not just problem-solving.
- The gap between speed and safety: Notably, Hassabis has long been a relatively cautious voice on AGI timelines, so his acceleration itself is a signal. He also stressed that winning does not mean running fastest, but stewarding AGI safely to everyone; safety infrastructure and international standards must keep pace with capability gains.
Original link: Read the original. This is a summarized translation for reference only.
Comments