MIT disavows doctoral student paper on AI’s productivity benefits

MIT says that as a consequence of considerations in regards to the “integrity” of a high-profile paper in regards to the results of synthetic intelligence on analysis and innovation, the paper needs to be “withdrawn from public discourse.”

The paper in query, “Synthetic Intelligence, Scientific Discovery, and Product Innovation,” was written by a doctoral pupil within the college’s economics program. It claimed to point out that the introduction of an AI device right into a large-but-unidentified supplies science lab led to the invention of extra supplies and extra patent filings, however at the price of decreasing researchers’ satisfaction with their work.

MIT economists Daron Acemoglu (who not too long ago won the Nobel Prize) and David Autor each praised the paper final yr, with Autor telling the Wall Street Journal he was “floored.” In an announcement included in MIT’s announcement on Friday, Acemoglu and Autor described the paper as “already recognized and mentioned extensively within the literature on AI and science, despite the fact that it has not been revealed in any refereed journal.”

Nonetheless, the 2 economists mentioned they now have “no confidence within the provenance, reliability or validity of the information and within the veracity of the analysis.”

According to the WSJ, a pc scientist with expertise in supplies science approached Acemoglu and Autor with considerations in January. They introduced these considerations to MIT, resulting in an inside evaluation.

MIT says that as a consequence of pupil privateness legal guidelines, it can’t disclose the outcomes of that evaluation, however the paper’s writer is “not at MIT.” And whereas the college’s announcement doesn’t identify the writer, each a preprint version of the paper and the preliminary press protection determine him as Aidan Toner-Rodgers. (TechCrunch has reached out to Toner-Rodgers for remark.)

MIT additionally says it has requested the paper be withdrawn from The Quarterly Journal of Economics, the place it was submitted for publication, and from the preprint web site arXiv. Apparently, solely a paper’s authors are purported to submit arXiv withdrawal requests, however MIT says “up to now, the writer has not carried out so.”