Thinker profile
AI policy researcher and independent scholar whose public interviews and writing appear in AI safety and governance discussions.
Profile basis
Capsule summary of public material and linked resources.
Ledger boundary
Prediction rows score claims, not the person overall.
Independent profile note
This page is an independent, AI-assisted editorial summary of public material about Carl Shulman. It is not affiliated with or endorsed by Carl Shulmanor any associated organisation. Biographical and reputational phrases are capsule summaries, not endorsements; use the linked sources, resources, and tracker notes for context, and read the scoring method before treating any scorecard as an accuracy ranking.
BA in Philosophy from Harvard University. Studied law at New York University School of Law. Born circa 1982; Canadian-American.
Research Associate, Future of Humanity Institute (Oxford, 2012–2024). Advisor to Open Philanthropy. Former Research Fellow, Machine Intelligence Research Institute. Former Director of Careers Research at 80,000 Hours.
Carl Shulman is an AI safety and existential-risk researcher whose public interviews and writing describe possible AGI-transition dynamics. His work covers the intelligence explosion, economic implications, geopolitical consequences, and alignment challenges. Public profile material describes him as an Open Philanthropy consultant and administrator of a reported $5M discretionary fund held by the Centre for Effective Altruism. His public writing discusses ideas including iterated embryo selection, donor lotteries, and hedonium. He has published on AI risk, decision theory, whole brain emulation, and the moral status of digital minds, including collaborations with Nick Bostrom and other Oxford researchers. While keeping a comparatively low public profile, his interviews and writing offer detailed arguments about AGI-transition dynamics.
4-hour marathon interview with Rob Wiblin. Covers intelligence explosion dynamics, economic growth post-AGI, 20-watt brain efficiency argument, the robotics bottleneck, training compute trajectories, and the physical automation constraint.
80,000 Hours Podcast · Episode #191 Pt 1 · June 27, 2024 · ~4h 15m
Listen / Read transcript →Covers AI-assisted governance, AI forecasting, locking in values, preventing AI-enabled coups, international treaties, the case against voluntary pauses, and binding controls at crunch time.
80,000 Hours Podcast · Episode #191 Pt 2 · July 5, 2024 · ~2h 21m
Listen / Read transcript →Long-form public interview covering how the intelligence explosion could happen before AGI, the fastest path to billions of robots, hardware fab constraints, and a step-by-step account of how AGI might attempt world takeover.
Dwarkesh Podcast · June 14, 2023 · ~2h 44m
Listen / Read transcript →Covers concrete AI takeover scenarios, bioweapons risk from advanced AI, cyber vulnerabilities, deception detection in AI systems, explosion speed as the key risk variable, and the 20-25% forcible takeover probability estimate.
Dwarkesh Podcast · June 2023 · ~4h+
Listen / Read transcript →Shulman's first long-form public interview. Makes the case that existential risk reduction passes standard cost-benefit analyses. Covers asteroid deflection analogy, expected value reasoning, and why the public health framing works for catastrophic risk.
80,000 Hours Podcast · Episode #112 · 2021
Listen / Read transcript →Co-authored with Elliott Thornley. Argues that standard cost-benefit analysis — not longtermism — already implies governments should spend much more on reducing catastrophic risk. Published in Essays on Longtermism.
Oxford University Press · with Elliott Thornley · 2025
Co-authored with William MacAskill, Aron Vallinder, Caspar Oesterheld, and Johannes Treutlein on decision theory under moral uncertainty — an altruistic agent uncertain between evidential and causal decision theory.
Working paper
Co-authored with Nick Bostrom. Explores the moral status of digital minds and how human-centric moral intuitions break down when applied to copyable, modifiable AI minds. Published in Rethinking Moral Status.
Oxford University Press · with Nick Bostrom · 52+ citations
Oxford Academic →Co-authored with Stuart Armstrong and Nick Bostrom. Models an AI arms race as a Nash equilibrium where teams skimp on safety to finish first. Finds that more information and more teams increase danger. 122+ citations.
AI & Society, vol. 31 · with Armstrong & Bostrom · 122+ citations
Springer →Co-authored with Nick Bostrom. Evaluates the claim that because evolution produced human intelligence, engineers should be able to create AI — and examines the selection effects that complicate the analogy.
Journal of Consciousness Studies, vol. 19, no. 7-8 · with Nick Bostrom
Shulman's personal blog covering AI risk, decision theory, effective altruism, immigration policy, and more. Sporadically updated and part of the public archive readers use to follow his arguments.
Visit blog →Extensive comment history and posts on LessWrong covering decision theory, Newcomb's problem, AI existential risk, and responses to AI governance proposals.
LessWrong profile →Contributions on effective altruism strategy, cause prioritization, donor lotteries, open borders, and wild animal welfare.
EA Forum profile →Comprehensive chronological list of all known Shulman publications, talks, and writing maintained by the Timelines wiki.
View timeline →