Agency and Complexity
Jonas discusses how altering rewards in reinforcement learning influences an agent's learning process, suggesting that interesting behaviors may emerge from maximizing expected free energy. Tim adds to this by exploring the balance between designing AI systems with human-compatible preferences and the desire for greater behavioral plasticity, drawing parallels to the complexity of human behavior shaped by DNA.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Jonas Hübotter (ETH) - Test Time Inference
Related Questions