Hacker News · AI·· 30 min agoAI score71
Astrophysicist uses AI to audit 158,059 papers in his field and finds 872 errors
I had AI read my field
AI summary
Casey Handmer, a Caltech-trained theoretical astrophysicist, used AI to screen 158,059 papers and 554,670 citations in his subfield, then had reviewers recompute 659 of the most-cited papers and verify 872 errors across 433 papers. Of those errors, 367 papers copied one, only 47 were ever criticised in print, and about half of copied errors were still being copied seven years after publication. The project cost roughly $2,720 in model time at list price, about $3 per verified error, and the deep check has so far reached only 4% of the highly cited papers on the map.
Source:Hacker News · AI · science.caseyhandmer.com