Lex Fridman avatar

Dexa/Lex Fridman

Learn more

Limits of Deep Learning

Noam discusses the distinction between engineering and science in the context of deep learning, emphasizing that while neural networks can identify patterns from vast amounts of data, they fall short of providing insights into the nature of human language. He critiques the reliance on random experiments in machine learning, arguing that true scientific inquiry requires purposeful experimentation. The conversation challenges listeners to reconsider the effectiveness of tools like Google parsers in advancing our understanding of linguistic principles.