This lecture provides a concise overview of building a ChatGPT-like model, covering both pretraining (language modeling) and post-training (SFT/RLHF). For each component, it explores common practices in data collection, algorithms, and evaluation methods. This guest lecture was delivered by Yann Dubois in Stanford’s CS229: Machine Learning course, in Summer 2024.
What's really happening inside the personal AI computer movement when everyone is defaulting to cloud models but the real power comes from owning the substrate underneath?
The common framing is local versus cloud — but the reality is that this is a routing decision, and the long-term reason to build your own stack is not cost savings but compounding your knowledge over time.
In this video, I share the inside scoop on how to build a personal AI computer that actually works:
• Why memory is the heart of the system and most people get the pipeline side wrong • How to set up many surfaces with one stack underneath so your editor, notes, browser, and voice all call the same runtime • What hardware makes sense for the local-first knowledge worker versus the all-local maximalist versus the local-first builder • Why cloud AI should be a visitor to your system, not dominant across it
Leaders renting their mem
