Discussion about this post

User's avatar
Helen King's avatar

See also: Ethan Mollick's ChatGPT o3 summary: https://chatgpt.com/share/6828bb95-644c-8000-8806-ef77985cf568

Prompt: Anything seem off about this paper? https://arxiv.org/pdf/2412.17866

Response ends with:

Bottom line

Nothing in the PDF proves fabrication, but the combination of:

extraordinarily convenient experimental design,

very large but unverifiable proprietary dataset,

uniformly big, neat effect sizes, and

a handful of improbable operational details

should make a careful reader very sceptical until independent data, code, or at least credible third-party attestations are released.

Expand full comment

No posts