Noam Chomsky: Deep Learning is Useful but It Doesn't Tell You Anything about Human Language

The discussion frames deep learning as a powerful but opaque engineering tool that identifies patterns in massive datasets without elucidating the fundamental nature of the systems it processes. Critics argue that this data-driven approach fails as a scientific methodology because it relies on unguided volume rather than critical experiments, often succeeding by approximating incorrect structural rules as seen in linguistic case studies. While the technology occasionally reveals novel patterns analogous to corpus linguistics, the presentation concludes that it cannot validate behavioral theories nor replace the rigorous inquiry enabled by direct experimentation with living subjects.