INTERACTIVE TOOL

P(doom) Calculator

"P(doom)" is the probability that advanced AI leads to catastrophe for humanity. It isn't one number — it's the product of several uncertain judgments. Move the sliders below to set your own view on each, and the calculator combines them into a single estimate you can compare against leading researchers.

1. AGI is actually built this century 75%

Probability that AI systems reach broadly human-level, strategically capable general intelligence before 2100.

2. Alignment isn't reliably solved in time 50%

Given AGI, the probability we fail to reliably align it with human intentions before it is deployed at scale.

3. A misaligned system gets real power 60%

Given unsolved alignment, the probability that a misaligned system is deployed in high-stakes, power-granting conditions.

4. It causes catastrophe, not a contained failure 50%

Given a powerful misaligned system, the probability the failure is civilization-scale rather than caught and contained.

5. Humanity cannot recover 70%

Given a catastrophe, the probability it is permanent — extinction or irreversible loss of human control over the future.

How this is calculated

The estimate is a simple multiplicative chain — each factor is the conditional probability of the next step, so the overall risk is their product:

P(doom) = P(AGI) × P(unaligned) × P(gets power) × P(catastrophe) × P(no recovery)

This mirrors the structure of Joe Carlsmith's 2021 report Is Power-Seeking AI an Existential Risk?, which decomposed the question into a chain of premises and multiplied point estimates together. The model is deliberately transparent rather than precise: its value is in making explicit which judgment is driving your number.

Note: a strict product assumes the steps are independent, which is a simplification — in reality these judgments correlate. Treat the output as a reasoning aid, not a forecast.

Where researchers land

AI optimists (e.g. Yann LeCun)
Public statements, various<1%
Median ML researcher
AI Impacts survey of 2,700+ researchers, 2023~5%
Joe Carlsmith
Open Philanthropy report, 2021 (later revised up)~10%
Paul Christiano
Alignment researcher, public estimates~20%
Geoffrey Hinton
Nobel laureate; interviews, 2023–2025 (10–50%)~50%
Eliezer Yudkowsky
MIRI; If Anyone Builds It, Everyone Dies>95%

Understand the number

Whatever your P(doom), the case for action holds.

Even a low probability of an irreversible, civilization-scale outcome carries enormous expected cost. That is the whole argument for treating AI safety as a serious priority now. Get our briefings on the risk and what can be done about it.

Keep exploring