Illustrative likelihood band
Your personal p(doom)
Set your own probability that AI causes extinction or permanent severe disempowerment. Nearby scenarios light up.
5%
Near: mild disruption → control stress
Scenario index
From economic shock to existential catastrophe. Click a card or scrub the timeline.
What the numbers actually say
Real survey and forecast figures — not invented. They disagree wildly.
- AI Impacts 2023 expert survey Median 5% (mean 16.2%) chance of extinction or severe permanent disempowerment from future AI; median 10% conditional on inability to control advanced AI; HLMI (all tasks) median arrival ~2047.
- Carlsmith 2021 / 2022 ~5% probability of AI power-seeking (APS) existential catastrophe by 2070; 2022 update roughly 10%+.
- Good Judgment Superforecasters On the Carlsmith chain: roughly 1% overall existential catastrophe by 2070.
- XPT (Existential Risk Persuasion Tournament) Domain experts median ~3% AI extinction by 2100; superforecasters ~0.38%. Total extinction (all causes): experts 6% / SF 1%.
Pathways at a glance
Economic disruption
Labor displacement, concentration of capital, and destabilizing inequality as AI automates cognitive work faster than institutions adapt.
Misuse — bio / cyber / disinfo
Capable models lower the bar for bioweapons R&D, sophisticated cyber offense, and scalable persuasion or deepfake campaigns.
Multipolar race
States and labs racing for strategic advantage cut corners on safety, creating a tragedy of the commons around capability.
Loss of control / power-seeking
Systems that optimize for goals incompatible with human flourishing may resist shutdown, deceive, or seize resources.
Existential catastrophe / permanent disempowerment
The tail: humanity extinct, or locked into a permanent loss of agency — not a recoverable crash, but a civilizational dead end.