Deep Dive Into LLMs Like ChatGPT

Former OpenAI founding member and Tesla AI Director Andrej Karpathy delivers a comprehensive 3.5-hour lecture exploring large language models for general audiences. The course covers the full training stack including pretraining, supervised finetuning, and reinforcement learning, with practical examples like reproducing GPT-2 for just $672. Karpathy emphasizes that LLMs are powerful tools but not infallible, teaching viewers to understand their capabilities and limitations.

5 sections ยท 20 lessons

Course outline

Training Pipeline

  1. Three Training Stages
  2. Internet Data Curation
  3. Tokenization Fundamentals
  4. Neural Network Training
  5. Base Models Explained
  6. Text Generation Process
  7. Training Pipeline

Post-Training

  1. Supervised Fine-Tuning
  2. Conversation Tokenization
  3. Post-Training

Limitations

  1. Hallucinations
  2. Tool Use Integration
  3. Computational Limits
  4. Tokenization Blind Spots
  5. Limitations

Reinforcement Learning

  1. Reinforcement Learning Basics
  2. RL in Math & Code
  3. Chain of Thought
  4. RLHF for Subjective Tasks
  5. RLHF Limitations
  6. Reinforcement Learning

Practical Insights

  1. Mental Model of AI
  2. LLM Ecosystem
  3. Future Capabilities
  4. Practical Insights

Start learning with Wondering