Researcher Nate Soares Estimates Over 50% AI Extinction Risk in Decade
Machine learning researcher Nate Soares estimates a greater than 50% chance of AI-induced human extinction within a decade, prompting him to halt retirement savings.
Machine learning researcher Nate Soares, a former software engineer at Google and Microsoft, estimates there is a greater than 50% chance that artificial intelligence will cause human extinction within the next decade. In a podcast interview with The Atlantic CEO Nicholas Thompson, Soares revealed that this belief led him to stop saving for retirement about ten years ago.
"It currently seems likely that we do not have a decade left," Soares stated, emphasizing that the probability of lacking a full decade is higher than 50%. This assessment represents his personal projection rather than a formal research finding. Soares maintains an existing retirement fund from his tenure at Google, but has not contributed to it in roughly a decade.
Soares currently serves as president of the Machine Intelligence Research Institute, a non-profit organization based in Berkeley known for promoting the pessimistic view that developing advanced artificial intelligence with superintelligence capabilities will lead to human extinction. During the interview, he compared living with this threat to living under the shadow of nuclear annihilation.
"This may be the first time we live under the possibility that all humans could die simultaneously, but on a personal level, we have always lived under the shadow of plagues and war," Soares noted. He added that individuals must do what is possible and continue living fulfilling lives, stressing that he does not view extinction as inevitable and works actively to prevent it.
Thompson expressed skepticism regarding the likelihood of AI-induced human extinction, noting that even the most advanced current technologies face substantial limitations. Soares countered this by drawing an analogy to the historical transition from horses to automobiles, suggesting that animals may have initially dismissed early motor cars for lacking certain capabilities while systemic improvements continued unchecked.
Clarifying his primary concern, Soares dismissed economic anxieties in favor of existential ones. "I am not a proponent of the 'AI job apocalypse.' I am a proponent of 'AI kills all humans,'" he remarked with grim humor, adding that employment rates would remain high right up until extinction. To mitigate this risk, Soares advocates for strict limitations on scaling the scale and capabilities of artificial intelligence systems, supporting beneficial medical applications like cancer research while opposing the pursuit of artificial general intelligence that matches or surpasses human cognitive abilities.
