Errata

No corrections to report yet. This page will be updated when errors are identified in the published text.

If you believe you have found an error, please contact us at smith@sfu.ca.

We will also post revised versions of key tables and figures, as new data comes in. We already have three updates to Table 8.1…

Table 8.1 Probability Estimates from AI Professionals and Others

Name Affiliation Credentials Estimate Date Notes
Geoffrey Hinton University of Toronto; Google Turing Award 2018; Nobel Prize Physics 2024 10-20% Dec 2024 “Gut” estimate; within 30 years
Yoshua Bengio Université de Montréal; Mila Turing Award 2018; chairs International AI Safety Report ~20% 2023-2025 Often frames qualitatively
Yann LeCun Meta AI Turing Award 2018 <0.01% 2023 “Less likely than asteroid impact”
Demis Hassabis Google DeepMind Nobel Prize Chemistry 2024 “Non-zero, non-negligible” Jul 2025 Declines to quantify
Dario Amodei Anthropic (CEO) Former VP Research, OpenAI 10-25% Sep 2025 Consistent since 2023
Sam Altman OpenAI (CEO) ~10-25% Various Less precisely stated
Elon Musk X, Tesla, SpaceX 10-20% 2024 Agrees with Hinton’s range
Shane Legg Google DeepMind Chief AGI Scientist; Co-founder DeepMind 5-50% 2011 Early estimate
Paul Christiano US AI Safety Institute Former OpenAI; founded ARC 10-20% 2023 “High enough to obsess over”
Eliezer Yudkowsky MIRI Theoretical AI safety >95% Various pessimistic prominent voice
Roman Yampolskiy University of Louisville AI safety researcher 99% 2024 Published academic
Dan Hendrycks Center for AI Safety ML researcher >80% 2023
Moulay Akhloufi Université de Moncton, N.B. Stanford University:“Among top 2% of AI researchers” existential threat from “Super AI” 2026 “real concerns”
Eric Schnidt Former CEO Google Senior Executive “Pull the Plug”? 2024 Catastrophic risks in future development
Paul Christiano OpenAI Board Board Member “4% over the next year and 15% over the next three years” 2026 rapid acceleration in capabilities leads to catastrophic and irreversible loss of control
Jacob Coxon OpenAI, Anthropic Quit AI Safety team in protest 10% within the decade 2026 People inside the companies are earnestly believing in extinction
Evan Hubinger Anthropic Supported Coxon’s statement >10% within the next decade 2026 “Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”

*Note for Paul Christiano: He estimates a 10-20% chance of “AI takeover” specifically, but a higher (50%) chance of “doom” shortly after human-level AI is reached.