AI Safety Technical Course 3 - RLHF: PPO & Agents

This lecture will cover PPO & Agents.
In this lecture:
Introduction to PPO (Proximal Policy Optimization) as a reinforcement learning algorithm and its advantages
PPO agent - defining the environment, policy and objectives
Learning phase - understanding how the agent improves through feedback
Training loop - putting everything together into a full RL training pipeline
Intoduction to RLHF: Connecting PPO to modern AI systems
To receive the certificate, you must complete the notebooks and attend this lecture, as in-person attendance is mandatory.

However, we will stream the lecture online for those who cannot attend in person.
Google Meet Link:
meet.google.com/kyu-utay-qia
Drift 23 is accessible through the library.

We'll be serving pizzas and snacks during the lecture.
After the lecture, you're invited to join us for drinks on the house.
Course material is inspired by ARENA, leading UK program in AI Safety.