Large Language Models (LLMs) are at the core of generative artificial intelligence (AI). To most of us, even those of us with technical backgrounds and expertise, they can seem like magic. In this talk, Ishan Anand will divulge the secrets behind how the LLM magic works using just a spreadsheet interface and some JavaScript and web components.

Inside the brain of an LLM: How it Works in 30 Minutes

Large Language Models like ChatGPT, Claude, & Llama may seem like magic but thanks to an innovative approach you don’t need a Ph.D. to understand how they work. In this session, we’ll use an innovative, web developer friendly approach to decoding AI using a spreadsheet interface built around JavaScript and web components. Using this fully functional implementation of GPT-2 (an early precursor to ChatGPT), we’ll break down the architecture of a basic LLM into digestible, interactive components that let you see inside the brain of an LLM as it works. In 30 minutes, you’ll walk through the anatomy of an LLM, understand how the basics of it works, and even control its mind without prompting.