Seminar by Prof. T. R. Ramadas, CMI on 11 November, 2026 at 2 pm

We are very happy to announce the following seminar by Prof. T. R. Ramadas. The details are as follows:

Abstract

Large Language Models (LLMs) are the mechanism behind conversational AIs (like ChatGPT, Gemini, etc.), coding agents like Claude Code, and recent spectacular (and disturbing) AI-generated proofs and counterexamples in Mathematics. An LLM is software that has been "trained" using large corpora of human text to make a choice of the "best next word" given a passage of text. I will explain how this is done.

This talk is based on a self-contained introduction to the mathematical architecture of LLMs:

A Self-Contained Mathematical Introduction to LLMs

This account should be accessible to anyone familiar with elements of linear algebra, multivariable calculus, and probability.