Experts debate wide range of AI extinction risk estimates

Commentators across academia, industry and the media have offered wildly divergent estimates for the likelihood that artificial intelligence could pose an existential threat to humanity, with figures ranging from virtually zero to more than ninety‑five percent. The debate has intensified as AI systems become more capable, prompting policymakers and researchers to grapple with how to assess and mitigate potential risks. Amid the noise, the numbers cited often lack a common methodological foundation, leaving decision‑makers without a clear basis for planning.

Columnist Jacob Aron, writing in a recent op‑ed, argues that the quoted probabilities should not be taken at face value. He points out that many estimates are derived from informal surveys, personal intuition or speculative modeling rather than systematic risk assessment, and that they frequently ignore uncertainties such as future regulatory environments, technological breakthroughs, and the complex dynamics of human‑machine interaction. Aron cautions that relying on such disparate figures can mislead both the public and policymakers, urging a shift toward transparent, evidence‑based approaches to evaluate AI safety concerns.