2 questions about htm theory

Community
, ,
1

  1. Is semantic folding theory the same as htm theory or a alternative but similar theory to htm or a theory that further built upon htm theory?

  2. Is the thousand brain theory a piece of htm theory or vise versa or are they both part of a broader theory of of the neocortex circuit ?

2
  1. I’m not familiar with semantic folding theory and I’ve not heard it mentioned in the contect of HTM.
  2. I don’t think we could call the thousand brain idea a theory in the formal scientific sense. It is an idea that has not yet been formalized. Maybe it could be called the thousand brain hypothesis but even then I don’t think it is making explicit claims of how it can be tested (I assume Numenta are working on this). HTM is a specific theory that associates macrocolumns in the neocortex with a computational model. It is a much “lower” level of detail than the thousand brain theory. I’m not sure the thousand brain theory requires the use of HTM, for example Numenta have been working with DNN recently. HTM seems to have been “stuck” with issues about how to create a scalable hierarchical system and Numenta have not published on HTM for many years, as far as I know.
3

  1. Semantic folding - Wikipedia

  2. The thousand brains theory says that each cortical area builds a fully functional model of every object that it receives sensory input about. Furthermore the theory says that each area of the cortex must be able to function independently from all other cortical areas. These two statements put strong constraints on any theory about how the cortical circuitry works. It also means that the cortex contains many independent models of each object, hence the name the thousand brains theory.

    HTM theory, is an older theory about of how the cortical circuitry works. It does not go into detail about how different cortical areas interact with each other.

1 Like
4

Yeah, define “object”, “fully functional”, and “independently”. There obviously are many orders of detail in objects and clusters thereof, and while cortical areas overlap in what they represent, they are obviously not fully redundant. So, I don’t see any strong constraints or predictions in TBT.

1 Like