Stanford's Noah Goodman uses philosophy to improve LLM pretraining, jokes ASI achieved
xuanalogue · x · 2026-09-26
Noah Goodman (Stanford) describes using philosophy to train LLMs: distinguish universal facts (true in all possible worlds) from contingent facts (true only in this world), let the model self-generate universal facts, and use them for pretraining.
He jokes he wanted to call the method Amortized Solomonoff Induction — meaning "ASI has been achieved at Stanford."
More from Research
- Position paper argues grounded abduction is the missing reasoning leap for LLMs — bibryam · 2026-09-26
- BeeNara: a 332MB CPU-only classifier that knows when no folder fits (96.8% recall) — razer_psycho · 2026-09-26
- InternLM open-sources Intern-Decision 4B/0.8B: structured decisions in one forward pass — jacek2023 · 2026-09-26
- Researchers surface spurious probes across models: Sonnet 5 recommends green tea in evals, oolong in production — jankulveit · 2026-09-26
- Three papers, one warning: 1% synthetic data can trigger strong model collapse — suchenzang · 2026-09-26
- Simulation beats distillation: the real story of synthetic data is post-training worlds — realsohamparekh · 2026-09-26