The P³ Index · Expert Panel
The same question, thirty-three answers, a 100-point spread.
The single most important fact about AI extinction risk is that serious, informed people disagree by enormous margins, from near zero to near certainty. This page places 33 leading voices, from industry, academia, policy, journalism, and this publication, on the P³ Index and lets you see that spread for yourself.
The spectrum
Each figure's own public "p(doom)" on a 0–100 scale. Ranges are shown as bars, single estimates as points. The dashed lines mark where thousands of working researchers land on average.
The panel
Probability is sourced. Possibility ("could it happen?"), Predictability ("would we see it coming?", higher means more foreseeable and controllable), and Concern are this project's interpretation, open to challenge.
Method & sources
Probability (sourced). Figures are drawn from public interviews, signed statements, and compiled trackers, including the community-maintained p(doom) reference and reporting by the BBC, CNN, Axios, and others. Where a figure gave a range (e.g. Hinton's 10–20%, Amodei's 10–25%), the range is shown. Where someone signed the 2023 extinction-risk statement but never gave a number (e.g. Altman, Hassabis), that is stated plainly rather than converted into a false precision.
Possibility, Predictability, Concern (interpretation). These translate each person's broader public position into a comparable 0–100 reading. They are editorial and revisable, the point of publishing them is to be corrected. Two reference points anchor the whole page: a 2023 survey of AI researchers found a mean extinction estimate near 14% and a median near 5%, and a 2024 survey of over 2,700 researchers again found a median near 5%.