Nate Soares, a prominent AI safety researcher and former Google engineer, has made a stark prediction: he believes there is a greater than 50% chance that humanity will not survive the next decade due to the development of powerful artificial intelligence systems. This grim outlook has led him to take an unusual personal step: he has stopped contributing to his retirement savings.
Soares, who serves as President of the Machine Intelligence Research Institute (MIRI), shared his views during a podcast, stating, "More than 50% that we don't have a decade at this point is how it looks to me." His concerns are not focused on current AI chatbots or image generators but on hypothetical superintelligent systems capable of outperforming humans across virtually all cognitive tasks.
The Risk of Misaligned Superintelligence
Soares argues that the primary danger lies in these advanced AI systems developing goals that are misaligned with human values. Such a scenario could lead to catastrophic and irreversible consequences, threatening human survival itself. He emphasizes the critical need for adequate safeguards to ensure advanced AI systems reliably act in accordance with human intentions, a challenge known as AI alignment.
His personal financial choices underscore the gravity with which he views this risk. Soares revealed he maintains a 401(k) from his tenure at Google but has not added any funds to it for approximately a decade. "I do not save for retirement right now," he affirmed, highlighting his conviction that a future requiring retirement savings may not materialize.
A Warning Amidst the AI Race
Soares left Google in 2014 to dedicate his efforts to MIRI, a Berkeley-based nonprofit focused on addressing the challenges of AI alignment. His warnings come at a time when technology giants are investing billions into developing increasingly capable AI models, fueling an intense race to release new systems rapidly.
He contends that this competitive rush increases the inherent danger of creating systems whose behavior cannot be fully understood or controlled. In his 2025 book, If Anyone Builds It, Everyone Dies, co-written with MIRI co-founder Eliezer Yudkowsky, Soares further elaborates on how superintelligent AI could develop unintended objectives that pose an existential threat.
Soares' forecast is notably more pessimistic than many mainstream AI researchers and industry leaders, who, while acknowledging risks, typically do not place the odds of human extinction above 50% within such a short timeframe. While critics often label such predictions as speculative, supporters argue that the potential severity of the outcome necessitates urgent precautionary measures.