Is 'AI is only good at verifiable domains' a cope that keeps holding up?
thinkingshivers · x · 2026-11-21
The author argues that the popular cope — 'sure, AI is great in verifiable domains like math and coding, but it's terrible in non-verifiable domains' — has actually held up quite well, and asks for the best reason to believe it won't remain true. The question highlights a core debate: without objective feedback signals, progress in creative, judgment-heavy, and open-ended tasks may lag behind.
Related event: Debate flares over claim that AI only excels in verifiable domains(2 posts)→
More from AGI Musings
- Denying a Fields Medal for AI-assisted RH proof would be like denying LIGO its Nobel — GiorgioPatrini · 2026-11-21
- Gary Marcus laments X has become clout-chasing and shitposting when AI needs better science — GaryMarcus · 2026-11-21
- Japanese tech lead argues code design matters more, not less, in the AI coding era — 4310sy · 2026-11-21
- Gary Marcus questions Anthropic's claim of discovering a new gene-editing mechanism — GaryMarcus · 2026-11-21
- AI ethics researcher rejects the ethics/safety divide, faults dogmatic voices — dioscuri · 2026-11-21
- Math professor: the real marvel isn't AI solving Riemann, it's humans building the machine — peter_richtarik · 2026-11-21