Thanks for clarifying.
Re. the DMN, yes I also agree that whatever the DMN reflects (modeling of internal states etc) is happening in each column. By scale I only meant that a Monty system would need to be large-scale for it to have system-wide activity that looked like the DMN in the brain - since the latter is by definition an observation of large-scale dynamics. Hope that’s clear.
Re. categorization: this is an interesting question and something you will generally here us discuss as a question of “categorization”, rather than using the words dimensionality (given the latter applies in so many settings). We are shifting the focus of our research to unsupervised learning (see the Future Work entries), and one thing we expect to observe is the merging of morphologically similar objects (e.g. different coffee cups) into single models. However, categorization is a complex thing with aspects beyond just morphology. For example, the commonality between coffee cups and wine glasses is more about affordances and their ability to hold fluids than it is about their morphology. Getting such grouping working is likely going to depend on us developing a better understanding of object behaviors, which is a work in progress.