Part 14/16:
He critiques the argument that neural networks like GPT-3 “don’t truly understand” because they lack human-like consciousness. Shapiro dismisses this as a no-true-Scotsman fallacy—assuming that human understanding is the only valid benchmark.
Instead, he advocates for a pragmatic view, understanding that knowledge emerges from evidence, beliefs, and consensus, not an absolute notion of truth. Humans, he suggests, are also limited in what they can truly know, relying on perception, interpretation, and social agreement.