Nearly 40% of new arXiv preprints are AI-written, raising training-data pollution fears
MaxUnfried · x · 2026-11-11
Cremieux's finding has replicated again: nearly 40% of new arXiv preprints appear to be AI-written. Max Unfried calls it 'recursive self-degradation' — AI slop science is raising the noise-to-signal ratio of research and polluting the training data for future LLMs.
More from AGI Musings
- 'Chat with your data' misses the point — people want a smart coworker — vboykis · 2026-11-11
- Ben Todd: AI task time horizons appear superexponential in AECI, not just revenue — ben_j_todd · 2026-11-11
- When AI becomes an oracle, what happens to mathematics? — stevenstrogatz · 2026-11-11
- David Patterson: superintelligence will end profit and equity growth, only sales-tax-funded universal high income works — davidpattersonx · 2026-11-11
- Even after a 6-month recursive leap, AI progress stays 3x the 2024 rate — ben_j_todd · 2026-11-11
- Forethought's calculator quantifies AI R&D automation speed-up without SIE — bookwormengr · 2026-11-11