This chapter explores the debate surrounding whether AI language models truly learn languages in a way that mirrors human learning. It discusses the contradictory views of experts and introduces the concept of 'impossible languages' as a testing ground.
AI language models were initially built without considering decades of linguistic research.
The models' ability to learn 'impossible languages' challenges traditional theories of language acquisition.
The chapter highlights the importance of understanding whether AI models learn languages superficially or genuinely.
Certain grammatical rules never appear in any known language. By constructing artificial languages that have these rules, linguists can use neural networks to explore how people learn.