Guardian: Even a Hiroshima-scale AI disaster might not wake humanity up

cccalum · x · 2026-08-24

Writing from Silicon Valley, the author finds experts expecting an extraordinary AI takeoff within two years—"recursive self-improvement," where AI trains each successive model, exponential growth toward superintelligence. A Stanford friend greeted him: "Welcome to the foothills of the singularity."

The quotes span the spectrum: Elon Musk predicts "an age of amazing abundance" while seeing a 10-20% chance of killer robots; Geoffrey Hinton puts p(doom) at 50% "because I haven't got a clue how to estimate the real number," adding "As soon as evolution of AI kicks in, we're fucked." The author's gloomy conclusion: perhaps nothing—not even a Hiroshima-scale AI disaster—will make humankind protect itself, and we may sleepwalk all the way to superintelligence.

Related event: Guardian: Silicon Valley Bets on AGI as Experts Sound Alarm(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →