Episode 24: Jack Parker-Holder, DeepMind, on open-endedness, evolving agents and environments, online adaptation, and offline learning
Generally Intelligent
World Models
The world models, sub-general, is under explored. It's because everyone has been so obsessed with this idea of learning with less data for RL. I think the reason why world models are really exciting is for generalization. You can train agents with Dynasile, all based RL. And you can also learn more about levels and unsupervised fashion. Do you feel like there are other underrated or overlooked papers or progenuse or techniques?
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
