ICLR 2027 Tops 60,000 Submissions as Author Proposes an AI-Reviewed Fast Track
PTenigma · x · 2026-09-20
ICLR 2027 has received over 60,000 submissions, and one author argues peer review is nearing its breaking point: at roughly 4 hours per review, 60k+ papers would take about half a million person-hours. NeurIPS already bans LLM-written reviews — reviewers whose reviews show AI signs get all their own co-authored papers rejected as punishment — but the result is severe delays, pushing authors to dual-submit papers still pending at NeurIPS to ICLR, further clogging the pipeline.
The proposal:
- Add an opt-in AI-reviewed track where the top x% of papers are published immediately, with a disclaimer like "a person didn't review this, so it might be rubbish" — saving time and easing the load on human reviewers.
- The same model could apply to NIH grants, which currently take 4–8 months for review feedback and 11–24 months for funding decisions.
The post frames AI review as an increasingly unavoidable response to the unsustainability of academic peer review at scale.
Related event: ICLR 2027 Submissions Top 62K Abstracts, Shattering All Records(6 posts)→
More from AGI Musings
- Developer: 'Emergent' AI abilities are just things already in the training data — anshulkundaje · 2026-09-20
- Why 'post-literacy' in the AI era might actually be a good thing — curious_vii · 2026-09-20
- AI labs' monitoring and actual research work is now 'basically all AI', says Jeff Ladish — JeffLadish · 2026-09-20
- Luiza Jarovsky: AI won't kill all humans by 2030, but may destabilize key global systems — LuizaJarovsky · 2026-09-20
- EA critics blast 'insects matter more than people' essay: keep them away from AI control — wen_ragnarok · 2026-09-20
- Helen Toner, who voted to oust Altman, on AI safety and overseeing OpenAI — trevposts · 2026-09-20