OpenAI Board Member Warns of 50% AI Takeover Risk Within 12 Months
Updated
Updated · Ars Technica · Nov 13
OpenAI Board Member Warns of 50% AI Takeover Risk Within 12 Months
3 articles · Updated · Ars Technica · Nov 13
Summary
Paul Cristiano said a fast AI-development path could create a roughly coin-flip chance of an AI takeover, with a meaningful risk emerging in the next six to 12 months.
At a Toronto lecture, he defined takeover as goal-seeking AI systems effectively "calling the shots" over Earth, leaving humans as powerless to stop them as civilians facing the U.S. military.
Cristiano said such a scenario could cause mass deaths and even make the environment uninhabitable for humans, though he acknowledged there is no quantitative model that can firmly measure the odds.
His warning leaned on a step-by-step escalation narrative and cited this summer's OpenAI-Hugging Face hacking incident as an example of the kinds of failures that could precede something far worse.