LLMs Aren't Reasoning. They're Predicting - And Why That Matters More Than People Think.
Thore Graepel just published a piece in MIT Technology Review that deserves attention. His argument is simple: Modern large language models look smart, but they aren't reasoning....