Gary King and Steven Worthington (Harvard) compared 134 reviews of 124 manuscripts by strong human reviewers with reviews by Claude Opus 4.8: three former journal editors preferred the AI review in 87% of comparisons, and on checking, 5.5% of the AI's claims about the manuscript were judged incorrect versus 15.9% for the humans
translated by AI, reviewed by the editors
· Science · Education
- technology
- Generative AI
- source
- gking.harvard.edu ↗
¶ chgzcd2kz