Niels recaps some of our key takeaways from the November focus brainstorming week (a week where we have these kinds of discussions every day for 4 hours each day). He dives into our ideas around learning causality, inverse causality, and using this for action outputs. We talk about which neural mechanisms and connectivity could be used to implement such functions and open questions around this. We also discuss how imagined states relate to real states and how imagined states can be used to express goals.
Summary Video
Main Video
0:00 Intro
2:09 Key Ideas from Mid November Brainstorming
8:22 Learning the Causal Action First
13:40 Behavior State Can Predict a Morphology State
35:28 Chain of Arbitrary Complexity - Examples
43:10 Rapid Learning in the Hippocampus
49:19 Observation: Learning to Play Piano
51:15 Learning Reverse Causality
52:30 Behavior Columns Diagram
54:58 Behavior Columns Diagram: Learning Forward Causality
1:00:50 Behavior Columns Diagram: Continuation
1:02:11 Behavior Columns Diagram: Initial Goal
1:05:49 Behavior Columns Diagram: “Bottoming Out” – Recruiting the Motor System
1:10:33 Observations on the Imagined States
