Paper says AI dominance is eroding its own innovation ecosystem

On October 21, 2026, a study titled "Stronger artificial intelligence breakthroughs hasten their own exhaustion" proposed that the dominance of AI research is eroding the innovation ecosystem on which its own breakthroughs depend. The study analyzed 7.18 million OpenAlex computer science papers, the cross-field mobility of 3.7 million researchers, and 73,089 peer reviews from 1990 to 2025, finding that from 2017 to 2025 the share of AI papers among computer science papers rose from 40% to over 55%, that 53.81% of 918,862 field switches flowed to AI, and that reviewers' confidence in their judgments of papers fell from 3.9 to 2.9 (on a five-point scale). The authors are from Tsinghua University, the University of Chicago, and the Santa Fe Institute, and argue that attention allocation, the review system, and reward structures need to reduce reinforcement of the currently dominant paradigm. The study has been reported by Unite.AI, and there are no further developments at present.

AI-generated from public reports

1. Oct 21· 1 reports

   Paper argues that AI dominance is eroding its own innovation ecosystem

   Source:Unite.AI