AI pioneer Geoffrey Hinton claims neural nets "understand" language better than human linguistic theory. Hinton, fresh off sharing a Nobel Prize in physics, doubled down on this in a recent interview with Nobel Prize Outreach’s Adam Smith.
Hinton said neural nets outperform the Chomskyan linguistics school, which is based on Noam Chomsky’s idea of innate universal grammar.
But not everyone buys it. A leading neuroscience language researcher with 20+ years of studying brain activity during language processing slammed the idea that AI truly "understands" language. 
The expert argues AI misses critical context humans use to make meaning. She highlights how emotional state, tone, and shared environment shape understanding, something current AI doesn’t grasp.
She also points out the difference between written text and real language. She used Hindi-Urdu and Serbian-Croatian examples to show mutually intelligible languages with different scripts.
The researcher cautions mixing up neural networks (algorithms) with real brain networks — and warns this confusion around AI’s "language understanding" risks dangerous misunderstandings.
Geoffrey Hinton’s Nobel Prize interview:
“What’s really surprised me is how good neural networks are at understanding natural language —
that happened much faster than I thought”….
“And I’m still amazed that they really do understand what they’re saying.”“Neural nets are much better at processing language than anything ever produced by the Chomskyan school of linguistics.”
Regardless of where a child is born, the human brain can acquire any language. (Unsplash/tommao wang)
