VC Stewart Alsop: LLM outputs never match intent — it's a desire problem, not technical
StewartalsopIII · x · 2026-08-26
VC Stewart Alsop argues he has never once prompted an LLM and gotten exactly what he wanted. He doesn't believe meta reinforcement learning will fix this, because it isn't a technical problem but a desire problem: humans project their wish to do no work at all onto the machine, expecting it to improvise perfectly on its own.
More from AGI Musings
- AI Transparency Exceeds Expectations: 2025 Prediction Review — MariusHobbhahn · 2026-08-26
- Model capability likely to dominate media in coming weeks amid intelligence jumps — iruletheworldmo · 2026-08-26
- Peter Diamandis: Today Is the Slowest Rate of Change You Will Experience — PeterDiamandis · 2026-08-26
- Creator laments: AI assumed in art, stealing human authenticity — MannerNo7000 · 2026-08-26
- a16z Partner: Distinction between training and inference will become meaningless — Kyrannio · 2026-08-26
- Audrey Tang tells NHK what people truly need in the age of AI — AnnaCiaunica · 2026-08-26