AI Takeover Could Happen Within Months, OpenAI Board Warns

Paul Christiano has spent years thinking about what happens when AI systems get smarter than the people trying to control them. Now the OpenAI Foundation board member and Alignment Research Center executive director is putting a timeline on it, and the timeline is short. At a lecture this week in a Toronto convention center, Christiano told the audience that an AI "takeover" of society, not a metaphor, an actual loss of human control over what happens on Earth, could be months rather than decades away.

This isn't a fringe figure making noise for clicks. Christiano sits on the board overseeing OpenAI's nonprofit mission and runs an organization dedicated to figuring out how to keep advanced AI systems aligned with human interests. His talk was the first in a three-part series laying out exactly how he thinks things could go wrong, how fast, and what it would actually look like if humans stopped being the ones calling the shots.

He was careful to note he has no hard quantitative model for any of this. That didn't stop him from giving the room a blunt estimate anyway.