Gary Marcus shares a thread tracking AI macro predictions that failed
GaryMarcus · x · 2026-08-04
Gary Marcus points to a live thread collecting AI macro predictions that turned out to be wrong, and invites people to add their own examples.
It is essentially a meta-discussion about overconfident long-range claims in AI and which forecasts failed in practice.
More from AGI Musings
- An AGI that cannot handle a four-way live conversation is too watered down, says one post — alexisgallagher · 2026-08-04
- Censorship could make permission matter more than intelligence — PierceLilholt · 2026-08-04
- Toby Walsh to launch new AGI risk book in Mosman on September 11 — TobyWalsh · 2026-08-04
- Math still matters because theorem proving trains the kind of reasoning AI-era life needs — jxmnop · 2026-08-04
- Misread alignment debate centers on whether GPT-3/4 work was already relevant to AGI — geoffreyirving · 2026-08-04
- AI-written ideas should be disclosed, not passed off as your own — 0xSammy · 2026-08-04