RL alignment may be unbeatable — pretraining is just a degenerate special case of RL
QuintinPope5 · x · 2026-10-13
Quintin Pope argues that deep learning's deep lesson is that training on data is by far the most effective way to instill particular tendencies or capabilities, which makes RL-based alignment hard to beat — there may be nothing else. He reframes pretraining itself as a degenerate special case of RL: offline, with fixed, equal imitation strength for every trajectory. On this view, "RL with appropriate data" subsumes every data-based method of instilling behaviors in models.
More from AGI Musings
- We taught machines to work like humans. Now AI measures humans like machines. — CurieuxExplorer · 2026-10-13
- In the AI era, knowing whom to listen to matters as much as what to build — jonathan_wilke · 2026-10-13
- Cathie Wood: AI could 5x global growth to ~15%, echoing Industrial Revolution — rohanpaul_ai · 2026-10-13
- Engineer warns AI-era tax rules may get even more convoluted since "the AIs can handle it" — MikePFrank · 2026-10-13
- IMF Forecast: AI Adds No Measurable Productivity Yet, Bust Would Cost Twice the Boom's Gain — rohanpaul_ai · 2026-10-13
- Timnit Gebru slams 'superintelligence' talk as unfalsifiable fictional delusion — AlexTensor · 2026-10-13