How to define P(doom) and why it matters

Christopher King, writing on LessWrong, tackles a tricky definitional problem: how do you even measure P(doom), the probability of existential catastrophe from artificial intelligence? Standard frameworks like betting or forecasting break down in apocalyptic scenarios, so King proposes a thought experiment instead. Imagine a genie offers to spin a wheel that randomly determines the world's fate—would you accept it? Your threshold for accepting or rejecting the wheel at different probabilities defines your P(doom). The piece illustrates how this illuminates disagreement among AI figures: Yudkowsky would spin for even a sliver of hope, LeCunn would reject it entirely, and Amodei sees some scenarios where the gamble might be worth it—each choice revealing their actual beliefs about AI risk.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton