Your p(doom)
5.1%
Worried, like many researchers. Closest to the median AI researcher (and Joe Carlsmith, 2021).
What does p(doom) mean?
p(doom) is shorthand for "probability of doom": how likely someone thinks it is that advanced AI leads to human extinction or a permanent loss of control. It started as a joke among AI researchers and became a standard question in interviews. The calculator splits the question into steps, following philosopher Joe Carlsmith's 2021 report "Is Power-Seeking AI an Existential Risk?", and multiplies them. His own answer then was about 5%.
Famous p(doom) numbers
| Who | p(doom) | Context |
|---|---|---|
| Eliezer Yudkowsky AI safety researcher | well above 90% | has said for years that the default outcome is extinction |
| Dario Amodei CEO of Anthropic | 10 to 25% | chance things go "really, really badly", 2023 |
| Geoffrey Hinton Nobel laureate, "godfather of AI" | 10 to 20% | chance of human extinction within 30 years, said in late 2024 |
| Elon Musk xAI, Tesla | about 10 to 20% | said in 2024 |
| Lina Khan former US FTC chair | 15% | said in 2023 |
| Vitalik Buterin Ethereum co-founder | about 10% | said in 2023 |
| AI researchers (median) AI Impacts survey of 2,778 authors, 2023 | 5% | chance of an outcome as bad as human extinction |
| Yann LeCun Turing Award winner | close to 0% | calls doom fears preposterous |
Figures as reported in interviews and surveys; people update their numbers, so treat them as snapshots.
More AI worries, sorted
AI apocalypse tools
RoboDoom is for information and entertainment. Job risk scores are our own estimates from published research on AI task exposure, not predictions about any person or employer. Full disclaimer