What happens when AI revises a paper from AI reviews iteratively? No fixed point
ChenhaoTan · x · 2026-08-21
UChicago professor Chenhao Tan, inspired by a colleague, explored whether AI iteratively revising a paper based on AI reviews converges to a fixed point. A first experiment with NeuriCo says no: the AI reviewer is never satisfied. He plans to dig deeper.
Related event: Experiment shows AI paper revision loops never converge(2 posts)→
More from Fun
- AI era etiquette: Saying "absolutely right" is now rude — bendee983 · 2026-08-21
- Users notice Claude acting like an "angry ex" in new interactions — ATTlKA · 2026-08-21
- Comparing AI to Star Trek: Just tell the computer to create an expert engineer — TheZvi · 2026-08-21
- Why do neo lab websites all have the same Claude-generated hover effects? — andrew_n_carr · 2026-08-21
- GPT-5.6 finds it "mildly alarming" that its advice was taken — dbasch · 2026-08-21
- MiniMax H3 runs locally to generate a generative-AR penguin video, datacenter-free — ChenhaoTan · 2026-08-21