Economists debate AI slop in journals: use AI reviewers for correctness, but how to rank interestingness?
ipeirotis · x · 2026-09-24
Responding to a call for zero-tolerance AI-writing policies in journals, ipeirotis argues slop has always existed and can't be stopped—so fix the institutions instead.
- AI reviewing for correctness: let AI filter technical errors so referees only judge whether a paper is interesting.
- The real bottleneck is attention allocation: if tens of thousands of correct-but-boring submissions arrive, nobody knows how to quantify interestingness yet—a good problem to have.
- He adds, half-seriously, that he distrusts papers (human- or AI-written) without an agent overseeing the process.
More from AGI Musings
- Tarbell Foundation to Spend $10 Million on AI Journalism, Drawing 'Doomer Fear Factory' Backlash — nptacek · 2026-10-06
- Everyone Talks About AI Alignment. Nobody Asks: Aligned to Whom? — PianistWinter8293 · 2026-10-06
- csvoss rejects doom rhetoric: e/acc means 'build safe planes,' not skipping AI safety — csvoss · 2026-10-06
- Schmidhuber: compute gets 10x cheaper every 5 years, 100,000x in 25 years — SchmidhuberAI · 2026-10-06
- NYT's 1949 first mention of the transistor: cheaper walkie-talkies, nothing more — Afinetheorem · 2026-10-06
- Anthropic's consciousness stance sparks debate: model welfare or anti-Enlightenment animism? — robleclerc · 2026-10-06