Experts are confident that they've figured out how to build artificial superintelligence. They're also confident that if they build it, the following outcomes are possible — and they've decided to put it to a global vote. It only gets built if the majority of people vote yes.
90%
Utopia — all diseases cured, people can live as long as they want, everyone has more wealth than the richest person alive today.
10%
Extinction — every human being on Earth is killed.
Why this is not purely hypothetical
Leading AI researchers and CEOs have publicly discussed non-trivial catastrophic risk from advanced AI, with some estimating existential risk in the 10%-20% range.
The same people building frontier systems increasingly suggest that AGI-level or "powerful AI" capability could arrive within years, not centuries.
Show quotes
Quotes are included for context and may be lightly edited for brevity. Source links are listed below.
- Geoffrey Hinton: “I think it’s somewhere between 10% and 20% chance that it will wipe us out.”[1]
- Yoshua Bengio: Bengio gave an estimate around 20% probability of catastrophic outcomes (p(doom)).[2]
- Dario Amodei: Amodei has discussed p(doom) in roughly the 10%-25% range and explicitly addressed existential-risk scenarios.[3]
- Elon Musk: “It’s most likely going to be great and there’s some chance — which could be 10% to 20% — that it goes bad.”[4]