AI生成の研究論文がピアレビューに氾濫
性能が向上したAIモデルにより、論文工場(paper mill)が説得力のある科学研究を大量生産できるようになり、検出できない編集者や査読者は対応しきれなくなっている。
researchers at the university of zurich traced a surge in citations to a 2017 paper back to ai-generated studies on the global burden of disease dataset. they discovered a chinese company advertising tools that allow users to churn out specific predictions, such as stroke likelihood or cancer rates, using large language models with minimal effort. while the papers contain methodological errors, they are sufficiently sophisticated to evade current screening mechanisms that previously relied on detecting obvious ai hallucinations. experts warn that generative ai is scaling the operations of paper mills, creating a crisis of reproducibility and burdening the limited supply of human peer reviewers.