Functionalists vs Landauer: why MoE sparsity analogies may extend to brains and consciousness
JoshPurtell · x · 2026-10-06
Josh Purtell argues in a debate with Rob Bensinger and others that functionalists should reason about the entire computational graph rather than instantaneously active complexity. He draws an analogy from sparse MoE models: a 112B-A3B model is generally stronger than a 3B dense one, so any Landauer-principle-based objection to machine consciousness would need to show a meaningful sparsity ratio between total brain complexity and instantly active complexity — which he says nobody has provided.
Related event: Functionalists Invoke MoE Sparse Activation Analogy in Consciousness Debate(2 posts)→
More from AGI Musings
- AI investor Derya pushes back on doomers: humans will be super-empowered, not ants — DeryaTR_ · 2026-10-06
- Nobody Needs a Personal AI Assistant — Freeing Up Time Just Means More TikTok — 0xAllen_ · 2026-10-06
- Ex-OpenAI Safety Lead: Some at AI Companies Have Reflected Deeply, Others Not at All — Miles_Brundage · 2026-10-06
- Ofir Press: someday we'll tell kids we used to see the whole raw reasoning trace — OfirPress · 2026-10-06
- AI Employees Should Map DC's Red Lines, Not Just Follow Company Policy — NathanpmYoung · 2026-10-06
- Lee Kai-fu brokers Kazakhstan's school AI pilot: 10 schools now, 500 by April 2027 — kaifulee · 2026-10-06