Discussion about this post

User's avatar
Ben's avatar
May 20Edited

Thanks for reading my piece. I have a few responses.

1) I think you've misunderstood and mischaracterized my views on peer review and preprints. The title "it begins with a preprint, as usual" was not meant to be snarky or dismissive. In fact, later on in the piece (in section 4) I explicitly defend the preprint system. I haven't written a post for my current blog on this, but in other forums I've written at length on why I think the modern peer review and academic publishing systems are incredibly flawed.

2) You write that the author exhibits deep domain knowledge in materials science and AI architectures. I think that the opposite is true; they have very superficial domain knowledge in those areas, viewing AI use and materials science innovation through the lens of an economics PhD student. This is the main reason why I think people were able to catch the fraud so early: a computational materials scientist wrote to the MIT department with their concerns, triggering the investigation.

3) You write:

"How can this be? It requires significant effort and expertise to pull this off and we are supposed to believe a 26 year old would invest so much effort in producing fake output. This doesn't add up."

What doesn't add up here? It adds up perfectly well to me. He got caught up in a chain of lies, wrote a paper with fake data (likely much of it AI generated itself), and then got caught. I'm not sure what more you think MIT could have done, other than not getting fooled in the first place (which, again, hindsight is 20/20). You demand accountability from MIT, but they (MIT, Acemoglu, and Autor) did exactly what they should have done. Upon getting a credible complaint that this student's work was fraudulent, they conducted a fairly prompt internal review, and then announced publicly that they could not stand behind the work and asked the journal and arxiv server to take the work down. If you want a multiple page accounting of how they could have been fooled by fraudulent work of a graduate student, well... for that there are substack writers.

Expand full comment
1 more comment...

No posts