3 free lectures: how ChatGPT & Claude work

3 free lectures that explain how ChatGPT & Claude work

Watch them in this order: short to long. All three are free on YouTube, no sign-up.

• 1. Large Language Models explained briefly (3Blue1Brown, 8 min) Start here. A visual intro to what a chatbot actually does: predicting the next word, pretraining, transformers. https://www.youtube.com/watch?v=LPZh9BOjkQs
• 2. Building Large Language Models (Stanford CS229, 1 h 44 min) A guest lecture by Yann Dubois (Stanford PhD student). How a ChatGPT-like model is built: pretraining, then post-training (SFT and RLHF), plus how the data and the evaluation work. https://www.youtube.com/watch?v=9vM4p9NN0Ts
• 3. Deep Dive into LLMs like ChatGPT (Andrej Karpathy, 3 h 31 min) Karpathy was a founding member at OpenAI. The full training stack for a general audience, the model's "psychology" (why it makes things up) and how to get the best out of it. https://www.youtube.com/watch?v=7xTGNNLPyMI

Tip: these explain LLMs in general. Claude is built on the same idea, so everything applies to it too.

Follow @ugurbuilds for more free resources like this.