Andrew Ng at Stanford: AI aces tasks with answers, lags badly on judgment calls
stanfordnlp · x · 2026-11-12
Andrew Ng guest-starred at Stanford's agents class CS 329Z, hosted by Diyi Yang, and a student shared her key takeaways.
- AI races ahead on tasks with a right answer (code, math) but lags far behind on judgment calls (which feature to build, whether to hire now) — a bigger gap than most assume. Benchmarks mostly test the first kind because it's easy to score: we measure what's easy and draw conclusions about everything.
- The knowledge gap between labs and everyone else is overstated and closes in months; what takes a decade is walking into a hospital or bank and making it actually run better.
- Ng says he has essentially stopped writing code and doesn't miss it.
Related event: Andrew Ng at Stanford: AI's Judgment Gap Is Underestimated(2 posts)→
More from AGI Musings
- AI agent traffic reportedly grew 8x in just two months — Dr_Singularity · 2026-11-12
- Ten gems from Diogo's Code Summit talk: RLHF is part of the problem, software is back — TheTuringPost · 2026-11-12
- If AGI is uncontrollable, should value anchor to life's unfolding, not just humanity? — danfaggella · 2026-11-12
- AI models can prove anything, making 'mathematically trivial' ICLR reviews obsolete — nanjiang_cs · 2026-11-12
- Neuroscience heavyweights debate whether brain foundation models can actually explain the brain — patrickmineault · 2026-11-12
- LLMs and GLP-1s both scaled to their size via novel structured finance, argues investor — abhiadesai · 2026-11-12