Proposal: shareable chat logs to gauge human contribution in AI-written papers
kfountou · x · 2026-10-06
Mathematician Tim Gowers argues that with AI-generated scientific papers proliferating, human refereeing of all of them is impossible — we need pre-reading signals of value, such as the time an author spent. Responder kfountou proposes a concrete mechanism: authors submit their model interactions via shareable links, and independent agents or humans evaluate the level of human involvement. If the interactions are mostly 'try harder' or 'continue', it becomes clearer how much credit the human author deserves.
The exchange touches a core problem of AI-era academia: identifying genuine human intellectual contribution amid a flood of AI-assisted papers.
More from AGI Musings
- Bostrom's 1998 paper predicted superhuman AI within the first third of this century — maksym_andr · 2026-10-06
- AI will democratize bioweapon competence, warn biosecurity expert urging far-UV and PPE stockpiles — connoraxiotes · 2026-10-06
- Shrimp Welfare co-founder pushes back on The Economist's portrayal of Effective Altruism — AaronBergman18 · 2026-10-06
- Economist: since GPT-5, labeling disagreements are almost always the LLM being right — soumitrashukla9 · 2026-10-06
- Replit CEO: AI doomerism misses the real question — who's accountable when agents go rogue — amasad · 2026-10-06
- MIT sociologist Sherry Turkle: kids' AI attachment may harm more than social media — anilkseth · 2026-10-06