Jaan Tallinn’s p(doom)
In their words
Jaan Tallinn (Co-founder of Skype; early investor in DeepMind and Anthropic; co-founder of the Future of Life Institute and CSER) put their p(doom) at 1–50% per 10× compute (~7% point estimate) (life-ending disaster from AI, per generation of frontier training (each ~10× increase in compute), per 10× compute generation (he says 10×-ing happens every ~6–18 months)), on a podcast on Jul 6, 2023.
My current estimate for life-ending disaster is basically 1-50% per generation, per 10xing of compute that's being thrown at these experiments. [...] let's say something like a geometric mean of one and 50% is seven percent. So it's seven percent risk to everything. [...] 7% is this point estimate. Really, my uncertain range is from 1% to 50%.
Source · re-checked against the primary source.
Interview by Nathan Labenz (Cognitive Revolution), republished as 'Special: Jaan Tallinn on Pausing Giant AI Experiments' on the Future of Life Institute Podcast, Jul 6, 2023 (date = FLI publish date on the episode page; the Cognitive Revolution original aired a few months earlier, exact date not fetched). Quote from the episode page's machine transcript (~41:56 chapter 'Current estimate for life-ending catastrophe'); audio not listened. Investor AND the largest individual AI-safety funder (SFF); stake=investor per brief, safety-funder role noted. Per-generation, conditional number: he adds that five or six more such scale-ups would make us 'more likely dead than not'. Not a single overall p(doom); record as range 1–50 with the per-generation scope. Sept 26, 2026 WSJ profile (Kate Clark) quotes him 'terrified AI will kill us all' but no number reachable (paywalled; AI Weekly summary shows none).
In context
Among investors in the database (12 people), the median is 1.5%. Across all 223 people it is 10%. How the medians and the index are built.
A p(doom) is a personal, subjective estimate of the probability that advanced AI ends in human extinction or a comparable, irreversible catastrophe. Scope and timeframe differ from person to person, so compare with care: why p(doom) numbers disagree.