Justin R. LesterFounder & AI-Ministry Strategist, TheoXTech · Pastor. Professor. AI Strategist.

So What is a Large Language Model? I got chu!

16 Sept 2025 · 1 min

Watch on YouTube

1:10

Justin R. Lester explains what large language models are and how they work. LLMs predict tokens based on embeddings and a context window, and can be retrained, fine-tuned, or aligned differently. They hallucinate, so verify outputs and check citations.

Chapters

So what is a large language model? My goal here is to try to demystify and give you clarity on what some of the things that are going on in technology are. And you probably heard a lot about large language models. I want to give you clearly what is a large language model? It's not just ChatGPT, it's not just Claude, it's not just Gemini. It's not just that. A large language model is a lot more than that. LLMs can predict the next token, tiny pieces of text based upon everything before. From that comes skills fluency emerges, summaries, drafts, outlines, code, key ideas. So I want you to think tokens and embeddings or the coordinates of meeting in the context window, the desk space. And the model has to hold your prompt and documents. Retraining makes it a generalist. Fine tuning specializes it and alignment teaches us to follow instructions and stay in bounds. Now they do hallucinate. They finish patterns, not facts. So you have to anchor them with retrieval and citations and verify anything you search for. So do not take the first draft, but you can take the second and third one after you fine tune and train these models. That's what a large language model is. So AI is not ChatGPT, AI is not Gemini, AI is not Claude, AI is much bigger than this. And large language models are part of the larger picture of artificial intelligence. All right, like and subscribe for more.