Mo BitarMay 31, 202610m

i was wrong about ai

TL;DR

A 1958 cat experiment reframed AI for him: Bitar opens with Hubel and Wiesel's discovery that a neuron in the visual cortex responds to an edge at a specific angle, then ties that directly to convolutional neural networks and the idea that intelligence can emerge from tiny specialized detectors.
Transformers solved the part CNNs were bad at: He explains that convolutional nets work well for images because nearby pixels matter, but language needs long-range relationships, which is why the 2017 paper "Attention Is All You Need" and transformers changed everything.
The real story is not new ideas, it's scale and hardware: Artificial neurons date to the 1940s and 1950s, training methods were worked out in the 1980s, and today's AI boom came when chips could do trillions of operations per second.
He now takes Richard Sutton's "bitter lesson" more seriously: Methods humans think are elegant often lose to dumb, general systems that improve when you throw compute at them, which Bitar once saw as fake science but now sees as possibly fundamental.
"It's just autocomplete" no longer feels like a satisfying dismissal: He argues next-token prediction is not merely parroting frequent words, because models are penalized for being wrong in context and can only improve by fitting deeper regularities in how text and the world hang together.
The video turns from technical history into existential unease: What starts as an AI architecture explainer ends with a much more personal fear, that if "simple things scaled up" explains vision and maybe intelligence, then AI is not just coming for jobs but for the human need to feel spiritually central.

The Breakdown

A single cat neuron firing at the angle of a line becomes the setup for a bigger confession: Mo Bitar says he may have been wrong to dismiss AI as just a scaled-up hack. His argument is unsettling and simple, that both vision and modern language models may be built from stupid little operations repeated at massive scale, and that possibility now feels less like trivia and more like a threat to human specialness.