Philosopher dioscuri: missing concepts don't block thinking — no deep human/LLM distinction
dioscuri · x · 2026-11-12
Responding to iamtrask's account of transformer weights as layered concept detectors, philosopher dioscuri argues there is no deep human/LLM distinction.
His argument: most of our concepts are learned — Aristotle lacked hovercraft and quarks; Helen Keller lacked the taste of Marmite and the sound of a theremin. Missing some concepts doesn't stop you from thinking with the rest.
He adds that LLMs exhibit many higher-level capabilities, and describing them only by their ingredients is explanatorily impoverished — it misses the real patterns, in Dennett's sense, that appear at higher levels of description.
Related event: Scholars debate whether Transformer weights are concept detectors(4 posts)→
More from AGI Musings
- Physician rebukes Dario: biology data covers <1%, 10-year cure claim is hyperbole — GaryMarcus · 2026-11-12
- Gary Marcus mocks Dario's 'cure cancer in 10 years' claim: echoes of Theranos — GaryMarcus · 2026-11-12
- Stardock CEO Brad Wardell: AI is the worst thing to happen to my game studios — draginol · 2026-11-12
- curl faces proposal to ban LLM-generated code as anti-AI pressure hits open source — steipete · 2026-11-12
- Baidu's Robin Li on why surging token spend isn't translating into profits: companies replaced the motor without redesigning the factory — Baidu_Inc · 2026-11-12
- disembodied intelligence didn't transform the world — hands, tools and communities did — neil_chilson · 2026-11-12