Now that AI can solve problems, optimize for theory-building
burny_tech · x · 2026-10-11
The argument: AIs don't do theory-building well only because the skill is useless until problem-solving works. Now that it does, models should optimize for proof size over a family of problems rather than a single one, and bespoke theory-building should fall out. Also worth cultivating: a sense of what makes a good open problem, as quality open problems may be drying up.
More from AGI Musings
- SF Tech Week takeaway: AI buzz shifts from agents to the infra agents need to run reliably — May_F1_ · 2026-10-11
- Dev proposes COGITATOR as a functional replacement for the term "AI" — solyarisoftware · 2026-10-11
- AI Impacts press release headline itself was misleading, not just journalists — jessi_cata · 2026-10-11
- Musk's X Takeover interview: Starship catch attempt in 4-6 weeks, why SpaceX bought Cursor, SI safety — XFreeze · 2026-10-11
- Toward an Automated Science of the Mind: AI Enters Every Stage of Cognitive Research — burny_tech · 2026-10-11
- Beff Jezos on post-labor economics: 'the solution has always been capitalism' — beffjezos · 2026-10-11