← Back to live feed1 story
A Harvard study of 1,124 academic papers found that Claude Opus 4.8 reviews had a 5.5% error rate compared to 15.9% for human expert reviews, with former journal editors preferring the AI's reviews 87% of the time in blind testing. The findings, based on 134 human reviews from rapidPeer, suggest AI can now handle the first full review pass of expert critiques, reversing previous conclusions that AI should only assist human peer reviewers.
Sign in to suggest
Key sources
- SOURCE@rohanpaul_ai“Only 5.5% of the AI's claims about the paper were wrong, versus 15.9% for the human reviews.”x.com
- SOURCE@teortaxestex“«AI-generated article reviews are now better, by every measure we could construct, than reviews written by well-compensated human experts»”x.com
- SUPPORT@tunguz“Peer review has always been crap, but never more than now.”x.com
- SOURCEhuggingnewshuggingnews.com