Demis Hassabis now expects artificial general intelligence within two to three years. Earlier this year, he said the target was still three to five years away. The shorter clock matters because the pace of advanced AI is moving faster than he had recently signaled.
Hassabis works in London and says the middle of the night is his favorite time to work, when he can read, think deeply and really focus in the small hours of the morning. He also says, "I feel like it helps me think original thoughts."
AGI in two to three years
He defines AGI as a system with the full cognitive range of the human brain. That is the benchmark behind his timeline, not a vague claim about better software.
He also says, "Step one, solve intelligence. Step two, solve everything else." His view is that the goal of intelligence is to understand science, and he has described that as the point of the work.
Google DeepMind and Hassabis
Hassabis has spent years leading work at Google DeepMind, and the broader story around him now includes a leadership shuffle there, with a new role taking shape as the organization adjusts around him. The practical takeaway for readers is that the person making the AGI call is still tied to the group building it.
That also means his timeline is not a detached prediction from the sidelines. It comes from someone who is still inside the effort, which makes the shift from three to five years to two to three years the detail to watch.
AI risks and solutions
Hassabis says AI risks are profound, but he also says, "It’s not all disaster, and it’s not all nirvana" and "I’m very much in the solutions business." He adds, "If there’s a problem, how do we solve that with collective human ingenuity? And if you don’t have a solution, then you probably need to work harder."
He says winning means helping humanity steward AGI into the world safely and securely for the benefit of everyone. The open question is what specific technical progress led him to shorten his AGI estimate now.
When Demis Hassabis says, "It’s soon" and, "I’m updating nearer and nearer," he is not talking about a distant research horizon. He is saying the next few years may be the window when institutions have to decide how to use, govern and contain systems that are moving toward human-level breadth.
