Yacine resurfaces his year-old hot take: drop pretraining, just RL
yacineMTB · x · 2026-10-07
Yacine digs up his own year-old tweet arguing that AI should "drop the pretraining, just the RL" — a take that looked contrarian then and is now part of mainstream debate. He laments/celebrates having called it early.
More from AGI Musings
- Princeton researcher: OpenAI's math breakthrough will hit coding in 6-18 months — brianryhuang · 2026-10-07
- Rationalist community's P(Doom) groupthink is driving its unusual behavior — teortaxesTex · 2026-10-07
- Nathan Lambert launches Trillium Labs, a nonprofit for open post-training frontier AI science — Miles_Brundage · 2026-10-07
- Reddit user: I don't want an AI assistant, I want AI that operates my computer for me — RGrayEsq · 2026-10-07
- Musk bets every 5GW of added US power equals roughly 1% GDP growth — XFreeze · 2026-10-07
- "A lifetime of open problems closed in one batch job": researchers' grief in the AI era — repligate · 2026-10-07