1. News
  2. AI models & agents
  3. AI · Timelines
AI · Timelines1 min read

DeepMind founder now expects human-level AI within 2 to 3 years

Demis Hassabis, a Nobel winner who built DeepMind, now says human-level AI could be two to three years away, down from his earlier guess of three to five.

By the Dashline deskSource: The Decoder
DeepMind founder now expects human-level AI within 2 to 3 years
Photo: Pauli Rautakorpi, CC BY 2.0

Demis Hassabis, the founder of Google DeepMind, now expects human-level AI within two to three years. Earlier this year he said three to five. He told the Wall Street Journal, "It's going to happen soon. My estimate keeps moving closer to the present."

What he means by it

The term is AGI, short for artificial general intelligence. Hassabis defines it as a system that matches the full range of the human brain's thinking. He has a test for it. Could an AI trained only on science from before 1901 come up with Einstein's 1905 papers?

He has described the impact as "10 times the industrial revolution at 10 times the speed." Hassabis is a Nobel laureate in chemistry. He has stepped down from Google DeepMind and now focuses on Isomorphic Labs, Google's drug discovery AI spin-off.

Safety and what is unknown

Hassabis recently pitched an oversight board that industry would fund and government would control. It would vet AI models before release. He also warns that bad regulation is worse than none.

This is one person's forecast, not a certainty. The material does not say how he reached the new number, how the board would work, or whether other AI leaders agree. For now, it is a reason to think about how your own work could change if he is even partly right.

The takeaway

A top AI builder keeps shortening his timeline for human-level AI, so it is worth thinking now about how your work could change.

More articles

You should really read these.

Next step

Let’s talk