Researchers Blast ML Conferences: Zero Downside for Bad Papers & Fraud
bremen79 · x · 2026-08-05
AI researcher Rylan Schaeffer criticized the fundamental problem with ML conferences: there is upside to doing bad work but zero downside. He noted that he has unearthed paper-invalidating problems with nearly every Best Paper and Oral he has touched, with two exposé papers forthcoming. kellerjordan0 added that most big lab researchers now read almost zero papers, understanding top venues like ICLR and NeurIPS to be mainly full of overclaims and fraud, with only a few diamonds in the rough.
Related event: Ex-Google Researcher Slams AI Conferences for Exaggeration and Fraud(6 posts)→
More from AGI Musings
- What Happens to Childhood When AI Educates with Perfect Precision? — PierceLilholt · 2026-08-05
- Miles Brundage: AI Companies Cut Corners Under Competitive Pressure — Miles_Brundage · 2026-08-05
- Top Medical Diagnosis Hit Rate is Only 35%, Highlighting AI Healthcare Potential — aakashgupta · 2026-08-05
- AI Pipelines Will Spark Academia's Worst Trust Crisis by 2027 — kevinnbass · 2026-08-05
- Rumor: Ilya's SSI Poised to Drop Breakthrough, Potentially Cracking Continual Learning — bindureddy · 2026-08-05
- The AI Bubble Disconnect: From Isolation to Superpower — RileyRalmuto · 2026-08-05