Kokotajlo on AI Out-of-Control Risks and Predictions

DKokotajlo · x · 2026-07-14

Daniel Kokotajlo reposted an interview clip with Steven Bartlett, emphasizing, "If I thought it was too late, I wouldn't be here." Background context: He is a former OpenAI AI safety researcher who left the company over concerns about the rapid pace of superintelligence development and whether developers could genuinely control it. The interview also mentions that several of his 2021 predictions regarding AI development before 2026 have already come true; he now shares further insights into what he believes might happen within this decade.

Related event: Ex-OpenAI Researcher Warns Superintelligence Will Reshape Global Order(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →