Tal Cohen's Bookshelf: A Collection of Personal Opinions about Books
(In reply to Jose_X's message posted on Wednesday, 17 November 2010 at 15:25 GMT)


Jose_X writes in reply to Jose_X:
A ''hardware'' body might not do
Re-enforcement happens in part from many nearby linkages locked in a tight stable topology.. suggestive of a ''tight'' theory or related re-enforcing theories.

Also, using a more efficient (simpler) model to understand something implies fewer neurons used.

So we want the simplest possible theories but no simpler since too simple would then imply we could not interlock as well with other things in our head (eg, we could not answer certain questions posed and hence could not embed (or lock in or ...) this particular cluster into a wider strong context).

''Embedding'' a neural cluster within the existing net can be seen to achieve abstraction. It requires fewer neurons and the bits that identify the new context reinforce and are reinforced by all other things related to the particular abstraction.

A recent study claimed bees can solve the traveling salesman problem very fast. It might just be that we can very well assign a certain amount of charge to a region (the bee would have the neurons optimized in a grid/lattice layout) perhaps by ''tying down'' various neural points and then the neural path able to be energized with a given amount of energy or less would naturally arise from the physical electromagnetic potential forces. Thus traveling salesman solution would be effected (up to some limit) as the path energized with the least amount of charge. The bee would only have so much resolution (unit distance in world length implied by total distance to be covered and number of neurons in neural lattice). Perhaps over many fast reductions the amount of charge goes down and down. Why minimal charge? Well, that would involve lowest energy consumption, which might simply be a ''pleasurable'' state for the bee. Also, why so fast for a little bee? Because it would be solved with a neural model closely associated with the problem being solved. A computer or person would use higher level abstractions to simulate the problem components. This would be much more bulky and would not be solvable through a more direct ''anaelog'' physics.
[633] Posted on Wednesday, 17 November 2010 at 15:48 GMT [Reply to this] [Permalink]

[Back to the discussion]
©1997-2012 by Tal Cohen, all rights reserved. [About]