AI Startups Criticized for Using Peer Review as PR Tactic

# research # iclr

A storm is brewing in the academic world over the use of AI-generated studies at this year's ICLR conference, a major event focused on artificial intelligence. Three AI labs, Sakana, Intology, and Autoscience, have stirred controversy by submitting AI-generated studies to ICLR workshops. Sakana took a transparent approach, informing ICLR leaders and obtaining consent from peer reviewers before submitting their AI-generated papers. However, Intology and Autoscience did not follow suit, submitting their studies without prior notification, as confirmed by an ICLR spokesperson to TechCrunch. The academic community has been vocal on social media, with many criticizing Intology and Autoscience for exploiting the peer review process. Prithviraj Ammanabrolu, an assistant professor at UC San Diego, expressed his frustration on X, highlighting the lack of consent from reviewers who provide their time and effort for free. He urged for full disclosure to editors about the use of AI in generating these studies. Peer review is already a demanding task, with a recent Nature survey indicating that 40% of academics spend two to four hours reviewing a single study. The workload is increasing, as evidenced by the 41% rise in submissions to the NeurIPS conference last year, totaling 17,491 papers. The issue of AI-generated content in academia is not new, with estimates suggesting that between 6.5% and 16.9% of papers submitted to AI conferences in 2023 contained synthetic text. However, using peer review as a means to benchmark and promote AI technology is a more recent development. Intology boasted about receiving unanimously positive reviews for their AI-generated papers on X, even quoting workshop reviewers praising the "clever ideas" in one of their studies. This self-promotion did not sit well with academics. Ashwinee Panda, a postdoctoral fellow at the University of Maryland, criticized the lack of respect shown to human reviewers by submitting AI-generated papers without their consent. Panda noted that Sakana had approached her workshop at ICLR, but she declined to participate, emphasizing the importance of respecting reviewers' time and rights. Skepticism about the value of AI-generated papers is widespread among researchers. Sakana acknowledged that their AI made "embarrassing" citation errors and that only one of their three submitted papers would have met the conference's standards. In a move towards transparency, Sakana withdrew their paper from ICLR. Alexander Doria, co-founder of AI startup Pleias, suggested the need for a "regulated company/public agency" to conduct paid, high-quality evaluations of AI-generated studies. He argued that researchers should be fully compensated for their time and that academia should not be used as a free resource for AI evaluations.

Related article

Comments (31)