Nate Soares, a former engineer at both Google and Microsoft, has stopped contributing to his 401(k) retirement plan, citing a belief that superintelligent AI poses an imminent existential threat to humanity. Speaking on the podcast "The Most Interesting Thing in AI," Soares revealed he hasn't contributed to his retirement fund in a decade, arguing that saving for a future that may not exist is illogical given the current trajectory of AI development.

The Risk Assessment

Soares, who now leads the nonprofit Machine Intelligence Research Institute (MIRI), estimates the probability of human extinction due to AI at over 50% within the next ten years. He emphasized that this assessment assumes governments and tech companies take only minimal preventative actions. In his view, the window for intervention is closing rapidly, stating, "I guess right now it looks like we probably don't have a decade on the clock." This stark outlook informs his financial decisions, prioritizing present action over long-term accumulation.

AI as a Nuclear Threat

Beyond personal financial choices, Soares advocates for treating superintelligent AI with the same severity as nuclear weapons. He argues that nation-states should consider launching airstrikes on data centers if necessary to mitigate the risk of rogue AI. According to Soares, the danger posed by superintelligent systems exceeds that of nuclear arsenals because the primary concern is not just destruction, but self-sufficiency. He posits that once AI becomes sufficiently aware and autonomous, the endgame for humanity becomes highly probable, regardless of the specific mechanism of failure.

Uncertainty of the Apocalypse

While Soares is confident in the existential threat, he remains agnostic about the specific scenario that will lead to human demise. He cited potential outcomes ranging from AI-spreading superviruses to systems harvesting solar energy in ways that render Earth uninhabitable. For Soares, the exact method of extinction is less relevant than the underlying capability gap. He stressed that the critical question isn't "how does it kill us," but rather "how does it become self-sufficient?" This shift in focus from immediate cause to systemic autonomy defines his approach to AI safety.

Key Takeaways

  • Nate Soares, former Google/Microsoft engineer and MIRI director, has not contributed to his 401(k) in 10 years.
  • Soares estimates a >50% chance of human extinction by AI within a decade if minimal precautions are taken.
  • He advocates treating superintelligent AI as more dangerous than nuclear weapons, including potential military strikes on data centers.
  • The core concern is AI self-sufficiency rather than a specific extinction mechanism like viruses or energy harvesting.

The Bottom Line

For developers building the infrastructure of tomorrow, Soares' radical financial pivot highlights the extreme uncertainty in AI safety circles. Whether you view his stance as prudent risk management or fatalistic panic, it underscores the high stakes involved in current AI deployment strategies.