Humans can't be SFT'd, only RL'd: a analogy for why persuasion fails
oran_ge · x · 2026-09-03
The author observes that humans are hard to persuade and only change after trying things and seeing results — analogizing the brain as a high-level model that can't be updated via supervised fine-tuning, only via reinforcement learning from its own experience.
More from AGI Musings
- Is the goal of AI memory to mimic human memory, or to be better than it? — AnuranBuilds · 2026-09-03
- Ben Affleck or AI CEOs: who explains the future of generative AI better? — SnoozeDoggyDog · 2026-09-03
- AI research integrity debate: critics say ML lags far behind other sciences' standards — silver__tsuki · 2026-09-03
- Pedro Domingos: AI destroyed mathematicians' illusion of sitting on a higher plane — pmddomingos · 2026-09-03
- If everyone can build frontier models, what's the end game for billions invested? — notadithyabhat · 2026-09-03
- Why doesn't an AI report the all-pervasive darkness? A consciousness thought experiment — yeastsplainer · 2026-09-03