tszzl: Pretraining science is like steam engines before thermodynamics, slouching toward interpretability
tszzl · x · 2026-10-16
Researcher tszzl quotes Andrew McCalip's piece on John Wilkinson and Watt's steam engine to argue that pretraining science resembles how good steam engines were built before thermodynamics was understood — "science owes more to the steam engine than the steam engine owes to science" (L.J. Henderson). If Nick is right, he suspects, progress points toward more interpretability, with pretraining "slouching towards monosemanticity," and he guesses major interpretability breakthroughs would also advance capabilities. He adds that meeting Nick was central to his own OpenAI origin story.
More from AGI Musings
- ETH Zurich's LFD lets a frozen LLM teach drug-design generators natural-language goals — rohanpaul_ai · 2026-10-16
- Chamath: 70-80% of AI company revenue comes from other AI startups; Zitron's bubble math — SumitGup · 2026-10-16
- Personal shopping agents bumped her secondhand spending from $365 to $820 in 2 months — davidyin44 · 2026-10-16
- There's no formal route to frontier AI research: 19-year-olds are advancing tech from their basements — repligate · 2026-10-16
- 'Will you end humanity?' AI's candid answer becomes a clown meme — dermflork · 2026-10-16
- Stanford professor: AI-done homework scores high, students fail in-person finals — FinanceYF5 · 2026-10-16