Scott Alexander on AI skeptics' endless shell game of moving goalposts
teortaxesTex · x · 2026-09-16
A passage from Scott Alexander's latest piece is circulating: AI skeptics play a never-ending shell game — whenever AI does something dramatic (escaping sandboxes, solving Millennium Problems), they say "of course, no one ever denied that," then point to the next undone thing as the real test, promising to reevaluate a worldview they never will.
- Quote: no matter how fast the rollercoaster accelerates, even after the familiar world vanishes behind you, you're still inventing reasons why it doesn't count.
- teortaxesTex notes the piece also contains one awesome Jewish joke and calls it clear and well-written.
More from AGI Musings
- tszzl: Labs sitting on solutions must reveal them fast — holding back is like a wood fence against tides — RileyRalmuto · 2026-09-16
- OpenAI president Greg Brockman: Execution beats ambition — you win by blocking and tackling — rohanpaul_ai · 2026-09-16
- Jensen Huang calls AI doom 10% prediction 'made up', slams labs' failed forecasts — PTrubey · 2026-09-16
- Debate: Do companies begging for regulation always want to lock in first-mover advantage? — nwilliams030 · 2026-09-16
- Charting the Agentic Garden of Forking Paths: Gelman flags agentic eval pitfalls — RexDouglass · 2026-09-16
- Allen vs Lonsdale AI debate: doomsday risk framing versus data-center innovation — MatthewChang · 2026-09-16