A math-first theory of intelligence may be the wrong lens, says X poster
fkasummer · x · 2026-07-23
The post pushes back on a claim that intelligence theory should be justified with a model-theoretic stability and minimality argument. The author says pure math is valuable, but if it is your main lens for building a theory of intelligence, you’re likely to have a bad time.
They argue this is especially true if you’re working outside the “legacy stack” of intelligence as function approximation via loss optimization. The deeper issue, they suggest, is epistemic: something about the way we know and formalize intelligence may be off, and the explanation could involve limits of computation or formalization rather than a clean mathematical proof style.
Related event: Debate: Pure Math May Mislead Intelligence Theory(2 posts)→
More from AGI Musings
- A proof prompt should restate the problem, list traps, and call in adversaries — burny_tech · 2026-07-23
- ChatGPT adoption could make developer experience everyone’s experience — jxnlco · 2026-07-23
- Investor says Google’s Gemini and DeepMind are too central to lose compute priority — firstadopter · 2026-07-23
- AGI-era retreat idea: islands where AI is banned to force hardship and human relationships — SuperbRiver7763 · 2026-07-23
- Liang Wenfeng’s cash-flow-backed AGI bet is a business model few founders can copy — yongqianme · 2026-07-23
- Multiple LLMs on one project start treating collaboration structure as a core feature — repligate · 2026-07-23