Economist finds LLMs 'so bad' at generating referee reports, tries translating as workaround
paulnovosad · x · 2026-09-30
Economist Paul Novosad shared that he tried using an LLM to generate referee reports on his own papers, but found the model "soooo bad" at it — so bad he's considering translating his work to Spanish and back as a workaround.
The post is part of the same thread where he criticized others for using LLMs to write peer reviews, adding a first-hand data point that the current models fall well short of usable review quality.
More from AGI Musings
- Musk says AI is becoming essential, cites cases where AI read X-rays right after doctors erred — XFreeze · 2026-09-30
- Prediction: in an agentic world humans won't code, write, or draw — editors have no future — akbirthko · 2026-09-30
- Would you sell your memories to an AI company? A privacy consent dilemma — VraserX · 2026-09-30
- Dean Ball mocks new AI discourse tribe prioritizing 'digital minds hacking' over sci-fi risks — deanwball · 2026-09-30
- Epoch AI's Greg Burnham on measuring AI progress: from math olympiads to Navier-Stokes — TWIML AI Podcast · 2026-09-30
- Developer predicts nobody will talk about AI 'moats' in a few years — eherrerosj · 2026-09-30