Why neural nets have to be so big: most bits are textures and history
jd_pressman · x · 2026-09-05
jdpressman argues that reproducing scientific observations outside of math requires mostly textures, lore, and specific bits of our worldline's history rather than clean laws—which is why neural networks have to be so large. He is currently testing intuitions about mitigating causal Goodhart without costly empirical experiments.
More from AGI Musings
- Many mathematicians value prestige over truth, discussion on AI proofs notes — avt_im · 2026-09-05
- WSJ: We're entering the era of artificial general intelligence — israelavila · 2026-09-05
- After 8 months of digging, researcher says persona models fail in RL — BronsonSchoen · 2026-09-05
- LLM demos now need 3D and games just to expose imperfections, researcher observes — airesearch12 · 2026-09-05
- Delivery riders demand platforms open the AI 'black box' they blame for cutting pay — nordicinst · 2026-09-05
- Reddit debate: Codex has 25M active users — why do some still insist AI is useless? — AkindaGood_programer · 2026-09-05