Thanks for the questions.
Re. agent types
- It’s definitely true that these exist on a continuum, however there are some subtle points about each that inform how we approach them. These are discussed in the bullet points on “Distant Agent” and “Surface Agent” on this page - I’d be curious if this answers your question.
Re. policies
- That’s great to hear you’re excited about the model-based policies, yes decomposition is a very important thing indeed! If you haven’t already seen it, you would probably be interested to read the section on Decomposing Goals in our Future Work part of the documentation.
Re. the default mode network
- That’s an interesting question, I don’t think it’s something we’ve considered too much. An initial thought is that something like this might emerge naturally in a large-scale Monty system when attending primarily to internally generated information (as opposed to external sensory information). Like you mention, higher cognitive functions require that the brain learn a model of itself, and so attending to internal information would probably be a natural basis for this network.
Re. Euclidean distance and computational complexity
- Let me know if the discussion in the other thread has helped with this, but otherwise happy to discuss it more.