Intro to Large Language Models

One of the best parts of YouTube Premium is being able to run audio in the background while your screen is turned off.

I utilized this feature heavily this past weekend as I drove back from a long weekend of camping. I got sick shortly before we left, so I drove separately and met my family the next day.

On the drive back, I threw on this video and couldn’t wait to tell my wife about it when we met up down the road at a McDonalds.

If you are completely uninterested in large language models, artificial intelligence, generative AI, or complex statistical modeling, then this video is perfect to throw on if you’re struggling with insomnia.

If you have even a passing interest in LLMs, though, you have to check this presentation out by Andrej Karpathy, a co-founder of OpenAI.

Using quite approchable language, he explains how you build and tune an LLM, why it’s so expensive, how they can improve, and where these tools are vulnerable to attacks such as jailbreaking and prompt injection.

I’ve played with LLMs for a few years now and this video greatly improved the mental model I’ve developed around how these tools work.