VC's hard-won take: LLM advice is fluent but often wrong — you must push back to get value
StewartalsopIII · x · 2026-11-13
Investor Stewart Alsop III argues that LLMs' linguistic fluency makes their advice look better than it is: it's often non-linear, missing crucial context even when you provide good context. He warns users to be wary of the "genie in the bottle," but notes LLMs can still give good advice if you become "disagreeable" and push back rather than accepting suggestions at face value.
More from AGI Musings
- Gary Marcus declares November the beginning of the end of the AI bubble — GaryMarcus · 2026-11-13
- Beyond p(Doom): Dan Faggella proposes a 'cosmic' p(Bloom) framing for AGI — danfaggella · 2026-11-13
- After Humans Fade, Which AGI Future Is 'Better'? A Three-Way Debate — danfaggella · 2026-11-13
- AI Agents Are the New AI Companions — and May Hit the Same Retention Wall — vaibhavbetter · 2026-11-13
- Schmidhuber gives Oxford talk on recursive self-improvement, a theme he traces to 1987 — mmbronstein · 2026-11-13
- AIFlow_ML claims centralized AI is data harvesting, questioning SF labs' funding — AIFlow_ML · 2026-11-13