As models get smarter, they often get worse at explaining things
saurabh_shah2 · x · 2026-07-24
The post argues for an “anti scaling law”: as models get bigger and smarter, they often get worse at explaining or teaching things.
The author says this also happens with human teachers once they truly understand a subject: smarter people tend to use denser language, making simple questions harder to answer in plain terms.
More from AGI Musings
- AI agents need identity, credit and atomic settlement, not human buttons — LexSokolin · 2026-07-24
- Jerry Tworek says the era of evals is over after seven years at OpenAI — agihouse_org · 2026-07-24
- A post argues AGI will be solved by AI before humans — DevToD4 · 2026-07-24
- A Reddit thread imagines one AI agent spanning voice, phone, car, glasses and robots — Life-Moose2698 · 2026-07-24
- What human thinking should remain when machines take over part of the work — pixlpa · 2026-07-24
- Post argues AI should be open source because training data was used without consent — ZeroStateReflex · 2026-07-24