AI · 2h ago
Karpathy's LLM Deep Dive: From Base Models to Chat Assistants
Andrej Karpathy's video explains how base LLMs are stochastic text generators that require post-training to become conversational assistants. It covers GPU vs CPU training, model releases, and limitations like poor counting. The talk provides resources for software engineers to understand LLM internals.
Meridian48 take
While the video is a solid primer for engineers, it glosses over the environmental cost of training and the ethical implications of deploying these models at scale.
llm-explanationandrej-karpathy