causal

LLMs work on impossible languages, so reveal nothing about language

Large language models have a built-in property that cannot be modified: they work just as well for impossible languages as for possible ones, scanning massive data for statistical regularities to predict the next word; because they make no distinction between possible and impossible languages, they can in principle tell us nothing about the nature of human language and thought, just as a periodic table including all impossible elements would tell us nothing about chemistry.

causalpending

Speaker

Noam Chomsky

Evidence Quote

“it does exactly as well with impossible systems as with languages. Therefore, in principle, it’s telling you nothing about language.”

Source

Noam Chomsky— Conversations with Tyler
Created: 6/12/2026, 4:53:54 PM

Your notes

Loading notes...