AI Safety Fundamentals: Alignment

83 Episodes

1. Is Power-Seeking AI an Existential Risk?

   Published: 13/05/2023
2. Where I Agree and Disagree with Eliezer

   Published: 13/05/2023
3. Supervising Strong Learners by Amplifying Weak Experts

   Published: 13/05/2023
4. Measuring Progress on Scalable Oversight for Large Language Models

   Published: 13/05/2023
5. Least-To-Most Prompting Enables Complex Reasoning in Large Language Models

   Published: 13/05/2023
6. Summarizing Books With Human Feedback

   Published: 13/05/2023
7. Takeaways From Our Robust Injury Classifier Project [Redwood Research]

   Published: 13/05/2023
8. AI Safety via Debatered Teaming Language Models With Language Models

   Published: 13/05/2023
9. High-Stakes Alignment via Adversarial Training [Redwood Research Report]

   Published: 13/05/2023
10. AI Safety via Debate

    Published: 13/05/2023
11. Robust Feature-Level Adversaries Are Interpretability Tools

    Published: 13/05/2023
12. Introduction to Logical Decision Theory for Computer Scientists

    Published: 13/05/2023
13. Debate Update: Obfuscated Arguments Problem

    Published: 13/05/2023
14. Discovering Latent Knowledge in Language Models Without Supervision

    Published: 13/05/2023
15. Feature Visualization

    Published: 13/05/2023
16. Toy Models of Superposition

    Published: 13/05/2023
17. Understanding Intermediate Layers Using Linear Classifier Probes

    Published: 13/05/2023
18. Acquisition of Chess Knowledge in Alphazero

    Published: 13/05/2023
19. Careers in Alignment

    Published: 13/05/2023
20. Embedded Agents

    Published: 13/05/2023