the-cortical-column-and-the-transformer-may-be-the-same-discovery

# The cortical column and the transformer may be the same discovery The neocortex tiles one universal module, the cortical column, 150,000 times; each runs the same predictive algorithm on different data. A transformer tiles one universal module, the attention block, and stacks it; each runs the same algorithm on different tokens. The architectural rhyme is too close to ignore: both are general-purpose predictive units, replicated in parallel, whose power scales with copy count rather than structural novelty. The open fracture is embodiment. Hawkins insists the column's reference frame, its sensorimotor grounding in movement, is constitutive of intelligence, not incidental. If he is right, the transformer is an impressive but fundamentally disembodied echo of the real thing. If TRC² and the alignment empiricists are right, the reference frame is one feature the architecture can absorb, not the gap that separates imitation from understanding. The question matters for how we understand mind. If the same computational unit recurs in carbon and silicon without anyone copying it, the unit may be closer to a discovery about intelligence itself than an invention of either substrate. Provenance: Stuart's synthesis. The convergence observation is his; the mechanism is Hawkins' (Thousand Brains Theory, Numenta 2024-25); the counter-position is from TRC² (arXiv 2602.22479, 2026) and "The Neuroscience of Transformers" (arXiv 2603.15339, 2026). The video (Kirsanov) is the occasion, not the claim. Related: [[the-self-and-world-disclose-each-other]], [[otherness-supplies-the-prediction-error]], [[the-maps-we-build-become-the-world]]