LLMs improve unevenly across fields, but still work across many without training
emollick · x · 2026-07-29
Ethan Mollick says LLM progress is jagged across fields: writing may improve slowly, but the striking part is how effective these models remain across many disciplines without explicit training.
More from AGI Musings
- Poll of 1,007 likely voters shows 86% want an AI off switch — wfithian · 2026-07-29
- Reddit thread argues true AI intelligence may be impossible to fully make safe — Successful-Lie1603 · 2026-07-29
- CSET says AI companies are already using AI to build the next generation of AI — hlntnr · 2026-07-29
- If AI always knows the truth, what happens to human lying? — PierceLilholt · 2026-07-29
- A warning that better models could be catastrophic if they remain misaligned — daniel_c0deb0t · 2026-07-29
- Would you hand GPT-6 your inbox, bank accounts and codebase to save 20 hours a week? — VraserX · 2026-07-29