Researchers banter: will human-generated pretraining data stay scarce?
DimitrisPapail · x · 2026-11-14
UW professor Dimitris Papailiopoulos exchanged a light jab with researchers Sewon Min and Xingyu Dang about pretraining data: Min mused that "human-generated" pretraining corpora may remain limited, to which Papailiopoulos replied, "why would you need human-generated?" — reflecting ongoing debate over whether synthetic data can fully replace human text.
More from AGI Musings
- Project Zero overhauls disclosure policy: AI made the 90+30-day timeline insufficient — ifsecure · 2026-11-14
- Terence Tao: math has moved from proof scarcity to proof abundance — haider1 · 2026-11-14
- 'Messy Jobs' makes Bloomberg's best books of 2026, with praise from Orszag and Brynjolfsson — TaniaBabina · 2026-11-14
- Commentary: Regulating open source hands China the win in AI's open future — StewartalsopIII · 2026-11-14
- Stanford COLM 2026 paper: LLM-generated teaching content is far less diverse than human work — HariSubramonyam · 2026-11-14
- Many AI disagreements boil down to one question: is superintelligence real — binarybits · 2026-11-14