"Sampling Is Not Analysis": Why LLMs Can't Take Responsibility for Decisions
gerardsans · x · 2026-11-20
Engineer gerardsans argues AI is fundamentally a text sampler — frozen weights, next-token sampling — and that has hard limits:
- It cannot take responsibility for decisions; when users or the law ask why, you need a human with lived experience.
- Git history and implicit trade-offs never written down were never in the training data — not in the data means not in the sampling support.
- Sampling is not analysis: no access to what led to a decision, correlation only.
- Using a recipe-book analogy, he adds that coverage is never infinite — the moment you need information outside the book, you need a cook.
Related event: Developer argues AI is just a text sampler, not a human replacement(2 posts)→
More from AGI Musings
- AI can't replace developers: no coverage, no accountability, no expertise — gerardsans · 2026-11-20
- Dev's long-form rebuttal: scaling can't conjure consciousness, 'quantum woo' hijacks anti-AI wave — AndyMasley · 2026-11-20
- "AI Is a Text Sampler": Why LLMs Can't Own Decisions or Replace Accountability — gerardsans · 2026-11-20
- AI stocks down 53% in 5 months — faster than the dot-com crash, says Gary Marcus — GaryMarcus · 2026-11-20
- AI Answers Are Replacing Clicks, and the Open Web May Never Be the Same — DavidLinthicum · 2026-11-20
- Gary Marcus amps doubt: no evidence AI profits will shift to apps — GaryMarcus · 2026-11-20