P(doom) is a term in AI safety that refers to the probability of catastrophic outcomes (or "doom") as a result of artificial intelligence.[1] The exact outcomes in question differ from one prediction to another, but generally allude to the existential risk from artificial general intelligence.
Originating as an inside joke among AI researchers, the term came to prominence in 2023 following the release of GPT-4, as high-profile figures such as Geoffrey Hinton[2] and Yoshua Bengio[3] began to warn of the risks of AI. In 2022, a survey of AI researchers, which had a 17% response rate, found that the majority believed there is at least a 10% chance that our inability to control AI could cause an existential catastrophe.[4]
| Name | data-sort-type=number | P(doom) | Notes |
|---|---|---|---|
| Dario Amodei | 10-25%[5] | CEO of Anthropic | |
| Elon Musk | 10-20%[6] | Businessman and CEO of X, Tesla, and SpaceX | |
| Paul Christiano | 50%[7] | Head of research at the US AI Safety Institute | |
| Lina Khan | 15% | Chair of the Federal Trade Commission | |
| Emmet Shear | 5-50% | Co-founder of Twitch and former interim CEO of OpenAI | |
| Geoffrey Hinton | 10%-50% | AI researcher, formerly of Google | |
| Yoshua Bengio | 20%[8] | Computer scientist and scientific director of the Montreal Institute for Learning Algorithms | |
| Jan Leike | 10-90%[9] | AI alignment researcher at Anthropic, formerly of DeepMind and OpenAI | |
| Vitalik Buterin | 10% | Cofounder of Ethereum | |
| Dan Hendrycks | 80%+ | Director of Center for AI Safety | |
| Grady Booch | American software engineer | ||
| Casey Newton | 5% | American technology journalist | |
| Eliezer Yudkowsky | 95%+ | Founder of the Machine Intelligence Research Institute | |
| Roman Yampolskiy | 99.9%[10] | Latvian computer scientist | |
| Marc Andreessen | 0%[11] | American businessman | |
| Yann Le Cun | <0.01%[12] | Chief AI Scientist at Meta | |
There has been some debate about the usefulness of P(doom) as a term, in part due to the lack of clarity about whether or not a given prediction is conditional on the existence of artificial general intelligence, the time frame, and the precise meaning of "doom".[13]