Nate Soares
President, Machine Intelligence Research Institute (MIRI)
Trace a path from Nate Soares →About
Nate Soares (born 1989) is President of the Machine Intelligence Research Institute (MIRI), the Berkeley nonprofit that has researched the technical problem of AI alignment since before the term existed. He holds a BS in computer science and economics from George Washington University and worked as a software engineer at Google and Microsoft, and as a research associate at the National Institute of Standards and Technology, before leaving Google in 2014 to join MIRI as a research fellow — becoming its executive director shortly afterward. That same year he co-authored the paper that introduced the term "AI alignment" itself, and went on to write extensively on corrigibility, value learning, decision theory, and the power-seeking incentives that make smarter-than-human systems dangerous by default. In 2023, when MIRI shifted its focus from alignment research toward warning policymakers and the public about AI risk, he moved from executive director to president — setting the organization's overall vision and strategy while Malo Bourgon took over as CEO. In 2025 he co-authored If Anyone Builds It, Everyone Dies with Eliezer Yudkowsky, arguing that building superintelligence with anything resembling current techniques would kill everyone, and that international coordination to prevent it is necessary rather than optional.
Key Contributions
- Co-authored the 2014 paper that introduced the term "AI alignment," naming the problem of making increasingly capable AI systems behave as intended
- Served as MIRI's executive director from shortly after joining in 2014 until 2023, when he transitioned to President as the organization shifted toward public advocacy
- Wrote extensively on corrigibility, value learning, decision theory, and the power-seeking incentives that make smarter-than-human systems dangerous by default
- Co-authored If Anyone Builds It, Everyone Dies (2025) with Eliezer Yudkowsky, arguing that current AI development techniques would produce a superintelligence that kills everyone
- Worked as a software engineer at Google and Microsoft, and as a research associate at the National Institute of Standards and Technology, before leaving industry for AI safety research in 2014
Videos & Interviews
AI Emergency: The AI Labs Are Lying To Everyone, He Says 99% Chance Of Extinction | Roman Yampolskiy
A four-way debate, which the title does not say. Ed Zitron, Andrew McAfee, Nate Soares and Roman Yampolskiy each wrote a probability of extinction in an envelope; the headline takes the highest of the four and hangs it on one name. The spread is the substance — the host reads out 99% and a rounding-error zero within seconds of each other.
View Details
The Rise and Reckoning of AI | 2026 Isaac Asimov Memorial Debate
Neil deGrasse Tyson moderates the 2026 Asimov Debate at the American Museum of Natural History, and the panel is built so the disciplines have to answer each other: Latanya Sweeney of Harvard on data privacy, Kate Crawford of USC on AI ethics and the material infrastructure underneath these systems, Chris Callison-Burch of Penn on natural language processing, Cynthia Rush of Columbia on statistical theory, Nate Soares of the Machine Intelligence Research Institute on long-term safety, and Eric Schmidt, formerly of Google, on national security and geopolitics.
View DetailsConnections
Alex Turner
DebatedIndependent AI Alignment Researcher
In 2022 Turner published a concrete shard-theory plan for the 'diamond maximizer' problem, and Soares answered with 'Contra shard theory', arguing that values learned this way are unnatural and would come apart once the system grew smart. Turner replied that he was not proposing an awesome new alignment technique but 'proposing that we don't need one.' That is the dispute in a sentence: whether ordinary training can instill values that last, or whether alignment needs something fundamentally new.
lesswrong.com
Roman Yampolskiy
In conversationAssociate Professor of Computer Science, University of Louisville
They sat on the same side of a four-way debate recorded on this site, and the closeness hides a real difference. Soares argues that no one knows how to align a superintelligence and that building one now kills everyone, a problem of time and method; Yampolskiy argues the problem has the wrong shape, like a perpetual-motion machine, since a system that must never err across every interaction and every round of self-improvement cannot be controlled indefinitely. Both want a ban, and only one thinks more time would help.
youtube.com
