Gary Marcus says item 2 is still unproven and hallucinations are counterevidence
GaryMarcus · x · 2026-07-24
Gary Marcus pushes back on the claim that item 2 is “done,” saying he does not see the evidence for it and that the closest test he has seen failed. He adds that hallucinations count as counterevidence.
The point is less about a specific benchmark than about how to judge progress: without convincing evidence, capability claims should not be treated as completed.
Related event: Gary Marcus Calls for Evidence in AI Debates, Citing Empty Claims(3 posts)→
More from AGI Musings
- AI access may move toward federal licensing, KYC, and shared blacklists — ctjlewis · 2026-07-24
- A long essay maps how engineers should grow in an AI-first workplace — bibryam · 2026-07-24
- AI can generate anything to taste, but the post says culture still lives in shared context — matdryhurst · 2026-07-24
- Zvi says you can’t call a model misaligned without knowing its system prompt — TheZvi · 2026-07-24
- Preprint proposes ACI to measure how alignment changes LLM triage decisions — zakkohane · 2026-07-24
- AI changed how one World Cup fan followed matches, from tactics to multi-model predictions — Cueus · 2026-07-24