AI Safety Fundamentals: Alignment

83 Episodes

1. Constitutional AI Harmlessness from AI Feedback

   Published: 19/07/2024
2. Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback

   Published: 19/07/2024
3. Illustrating Reinforcement Learning from Human Feedback (RLHF)

   Published: 19/07/2024
4. Chinchilla’s Wild Implications

   Published: 17/06/2024
5. Deep Double Descent

   Published: 17/06/2024
6. Intro to Brain-Like-AGI Safety

   Published: 17/06/2024
7. Eliciting Latent Knowledge

   Published: 17/06/2024
8. Toy Models of Superposition

   Published: 17/06/2024
9. Least-To-Most Prompting Enables Complex Reasoning in Large Language Models

   Published: 17/06/2024
10. Discovering Latent Knowledge in Language Models Without Supervision

    Published: 17/06/2024
11. ABS: Scanning Neural Networks for Back-Doors by Artificial Brain Stimulation

    Published: 17/06/2024
12. Two-Turn Debate Doesn’t Help Humans Answer Hard Reading Comprehension Questions

    Published: 17/06/2024
13. Imitative Generalisation (AKA ‘Learning the Prior’)

    Published: 17/06/2024
14. An Investigation of Model-Free Planning

    Published: 17/06/2024
15. Low-Stakes Alignment

    Published: 17/06/2024
16. Gradient Hacking: Definitions and Examples

    Published: 17/06/2024
17. Empirical Findings Generalize Surprisingly Far

    Published: 17/06/2024
18. Compute Trends Across Three Eras of Machine Learning

    Published: 13/06/2024
19. Worst-Case Thinking in AI Alignment

    Published: 29/05/2024
20. Public by Default: How We Manage Information Visibility at Get on Board

    Published: 12/05/2024