Speaker: Paul Smolensky (Microsoft/Johns Hopkins University)
Title: Do the syntactic abilities of generative AI systems falsify fundamental principles of generative linguistics
When: Friday, April 24th, 3:30-5pm
Where: 32-141
Do the impressive abilities of neural-network Large Language Models in generating rich, well-formed syntax falsify fundamental principles of generative linguistic theory? The answer I will argue for is: no. But it will be a rather nuanced “no”, trying to identify the proper treatment of generative AI for generative linguistics. Specifically, I will consider these principles:
- Computability: Generating natural language with rich, human-level syntax requires use of symbolic grammatical rule systems.
- Explanation: Theoretical explanation in generative linguistics requires built-in discrete symbolic structure.
- Acquisition: Children’s ability to acquire language requires innate knowledge of grammatical rule systems.
- Universals: Linguistic universals can only be explained from innate limitations on what languages are learnable.
The quantity of discussion of these questions will decrease sharply from 1–4, the bulk of the presentation focused on 1. The discussion of 1 takes off from Smolensky, Fernandez, Zhou, Opper, Davies & Gao (JAIR 2025; arXiv:2410.17498).