TL;DR: AI written reviews, and AI-writing in papers
I was going to give this a title "An Obituary of the Review Process", but in the interest of not sounding too defeatist, I will just say "Critical Condition" because of AI writing, both on the author side and the reviewer side.
I am a reviewer for AAAI 2027 and ICLR 2027. Since Phase 1 and Phase 2 was completed, I can now see other reviews. I can bet good money that about half of the reviews in the papers I'm seeing are AI-generated completely, or edited after AI-generation. This is subjective opinion and I cannot prove it but I will try to convince you - at least convince you that it's a problem that needs much much more attention than is being currently given.
I have been reviewing for top conferences (NeurIPS, ICML, AAAI, ICLR) since 2022, in total about 10 times.…