
Episode 19: Minqi Jiang, UCL, on environment and curriculum design for general RL agents
Generally Intelligent
00:00
Learning to Communicate With Machines
The first in learning project that i did was actually implementing this paper called learning to communicate, which was by an old friend of mine. He looked into training neural networks using centralized training decentralized execution to solve communication riddles. I thought just thinking about those problems was really interesting. It is such an open ended process where if you have these adversarial curricula, you can really start have a system that starts to steer into whole new areas of the agent environment spacethat you just don't expect.
Transcript
Play full episode