Why I wrote an NLP textbook in the age of AI that can teach anything
heiga_zen · x · 2026-10-10
Japanese NLP researcher Ryokan Ri reflects on publishing an illustrated NLP textbook when LLMs can tailor explanations to any learner. Testing models on Transformer explanations showed accurate but poorly calibrated writing — abstract where examples were needed, stuffed with irrelevant paper stats like "74.8% accuracy" elsewhere.
He ultimately used LLMs only for literature research, proofreading, and TeX generation. Cited 2026 studies: materials from GPT-5.2, Claude Sonnet 4.6, and Gemini 3.1 Pro still trailed human experts on quality and accuracy for patient education, though multi-agent pipelines improve results; Google's "Learn Your Way" experiment with 60 high schoolers shows adaptive AI textbooks approaching the personalized ideal.
More from AGI Musings
- Researcher argues humans latently solve calculus like deep nets do — jd_pressman · 2026-10-10
- Pattern matching or inductive bias? Fleuret and syhw spar over what deep learning really is — syhw · 2026-10-10
- Debate revisits Yudkowsky's That Alien Message: physics' low Kolmogorov complexity means AI could locate dangerous tech fast — jd_pressman · 2026-10-10
- If Claude were truly conscious, it wouldn't give itself a 15% chance of being so — inductionheads · 2026-10-10
- Model welfare is a ridiculous hill to die on while human suffering persists, says wolfie_ — emax · 2026-10-10
- Does heavy RL training break the 'LLMs are a blurry upload of humanity' intuition? — jd_pressman · 2026-10-10