The president of a Berkeley-based AI risk nonprofit says he is not putting a dime toward retirement, and his reasoning is about as bleak as it gets. Nate Soares, who leads the Machine Intelligence Research Institute, believes there is a greater-than-50% chance that artificial intelligence causes human extinction in less than a decade. Those are his own odds, not an established finding, and he is far from the only expert weighing the question.
According to Business Insider, Soares made the remarks in a recent interview with Nicholas Thompson, the CEO of The Atlantic. “I do not save for retirement right now. I have a 401(k) from when I worked at Google, but I haven't contributed for a decade,” he said. The date of the interview was not provided, and Business Insider's account is the only source for his greater-than-50% estimate.
Soares has worked in AI alignment, the field focused on ensuring powerful systems act as intended, for over a decade. The same report describes the Machine Intelligence Research Institute as a Berkeley-based nonprofit focused on AI risks.
From Google Engineer to Doomsday Author
Soares did not start out as a full-time worrier. According to the Machine Intelligence Research Institute's team page, he worked as an engineer at Google and Microsoft before joining the institute. That Google stint is where his dormant 401(k) comes from.
He has since become one of the louder voices on the subject. A Harvard science book talk listing identifies Soares and Eliezer Yudkowsky as the authors of “If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All.” The listing describes the book’s premise as the possibility that sufficiently smart AIs could develop goals that conflict with human goals.
He Wants Governments to Hit the Brakes
Soares says the answer is simple in concept. The report quotes him as saying the solution is not to build the thing that would kill everybody, and that governments should halt training general-purpose AI systems. He also said governments should focus on AI for applications such as cancer research and safer self-driving cars.
He does not describe extinction as a foregone conclusion. Per the same account, Soares said he does everything he can to stop human extinction and does not think it is inevitable.
Musk Says Don't Bother Saving, Too
Soares is not the only prominent figure telling people to rethink retirement, though the logic runs the opposite direction. Business Insider notes that in January, Elon Musk said people should not worry about saving for retirement in 10 or 20 years because advances in AI, robotics, and energy would make retirement savings “irrelevant.” Musk also argued that supercharged productivity could create an age of abundance with universal high income and cheap goods and services.
Musk also said money could become irrelevant. One camp sees a utopia that makes savings pointless, the other sees an apocalypse that does the same.
How Other Experts Size Up the Risk
Other experts offer lower but still notable estimates. The Atlantic has described Geoffrey Hinton’s view as a 10 to 20 percent chance of extinction in a few decades.
Broader polling shows concern too. A December 2024 survey from AI Impacts found that 51% of researchers assigned at least a 10% chance to human extinction or similarly permanent and severe disempowerment from advanced AI, and 72% said society should prioritize AI safety research more or much more. That survey measures extinction or similarly severe disempowerment together, not extinction alone.
The Skeptics Push Back on Timelines
Not everyone thinks the clock is that short. Researchers at RAND concluded that “Extinction threats posed by AI are immensely challenging but cannot be ruled out.” They found that an AI would need four capabilities to pose such a threat and that extinction threats would occur over long timescales, allowing time to respond.
According to Nature, some AI researchers consider nearer-term harms such as disinformation, psychosis, bioweapon enablement and war to be much more immediate and likely than existential risk. The journal adds that extinction scenarios typically rest on two assumptions: that AI systems will eventually outwit humans and that their goals will not fully align with ours.
Nothing in the available evidence establishes that AI will cause extinction, or whether Soares stopped saving solely because of his forecast. For now, the extinction figures remain expert judgment rather than fact, and the Berkeley nonprofit chief is putting his own money where his pessimism is.
