I’ve finished reading Jeff Hawkin’s book “A Thousand Brains”, and It left me with questions. First of all, lets take reference frames such as one centered on a cup. What questions can we answer such a ‘reference frame’? We could identify the object, and predict the distance between two parts of the object and perhaps how part of the object moves relative to the rest (like a stapler). Now lets take more abstract reference frames - such as a map of words where words with different meanings are close together, and very different words are far apart. Again, what questions could we answer - just that two words are close together? In a mathematical reference frame - what would it mean to solve a problem - where would you move from the initial equation in “equation space” to the solution of the equation?
Suppose you had a goal that you represented in the cortex, and you also had a starting point from which you wanted to reach the goal. The example could be spatial, you represent your current position at home, and your goal position at the ice-cream truck outside. How do you solve the problem of getting from one position to the other with a sequence of positions in the cortex or hippocampus reference frame?
I don’t see how the theory leads to many mundane actions that we do. Maybe thats all in the interactions of the “old brain” with the neocortex?
I dont think this is yet knowable … TBT so far does not cover Reinforcement Learning and Planning.
CC receives Motor commands (ACTION) and sends back Location and Sense information i.e. STATE.
I would assume that this loop goes trough Basal Ganglia.
Planning probably involves hierarchy of CC’s.
How I imagine it … example…
If you try to imagine your drive from home to the grocery you can not imagine all steps at once you can do it only sequentially…
here is my speculation … higher level CC’s groks Home=>Grocery which down-propagates Home=>Landmark1, then inner loop recalls Lanmark1=>Landmark2, L2=>L3 … Lx=>Grocery.
I have a speculation that the Grid/RFrame make the physical non-linear distances become virtual linear locations i.e. I say step/landmark 1,2,3… but not 200m to L1, next 567m to L2, next … unless u need the specific data it is easier to recall/plan
At any point this process can be interrupted and modified via the BG loop!!
When the path is not known we rely only on BG loop…
Just throwing random thoughts … I’m also interested how this could work !!