
Models and Foundational Models Explained

AI is Smart Until It's Dumb: Why LLM Will Fail When You Least Expect It - Emmanuel Maggiori
Topic 9 of 19
Published Nov 10, 2025 · Clip 27:38–30:01 (2:24)
Emmanuel defines a model as the final result of training—a computer program template with parameters (blanks) filled in by a separate training algorithm, which is why the word 'program' is avoided to prevent confusion. Foundational models are very general-purpose LLMs that can do many things and serve as building blocks: apps can connect via API or fine-tune an open-source foundational model on their own data for specific problems.