Grok evaluates a BrainCode-vs-MindCode thesis: LLMs still differ from brains
wschroll · x · 2026-09-14
The author used Grok to evaluate his "BrainCode vs MindCode" thesis, arguing that Gary Marcus's neurosymbolic approach and Yann LeCun's world models are interesting but may not suffice. LLMs are highly successful in text and even abstract domains like mathematical proofs, yet they differ fundamentally from an organism's endless parallel loops across brain-body-environment, drifting attractors with slow neuronal switching, and minimal energy use compared to data-center training and inference.
More from AGI Musings
- Researcher bets he'd beat the entire alignment field in 4 weeks with a cabin and some GPUs — RileyRalmuto · 2026-09-14
- AI researcher: no real scientific consensus on AI safety, warns of interest-driven fearmongering — JJitsev · 2026-09-14
- Report: No 1M-GPU Vera Rubin run in 2027, but a 500K run next year; 15-25% compute earmarked for alignment — teortaxesTex · 2026-09-14
- There is no single "AI alignment problem": researcher offers a structured taxonomy of its many faces — Dr_Atoosa · 2026-09-14
- Nature: Scientists test whether LLMs trained on historical data can reproduce creative breakthroughs — TZahavy · 2026-09-14
- Nate Silver: AI x-risk arguments lack empirical grounding and rely on bad analogies — davidmanheim · 2026-09-14