Connectionists: Virtual concepts

[Stephen Jose Hanson]

Hi Richard,

you raise many good points which I agree with and agreed with when I 
read and edited your chapter.

(Btw the actual Reference is :   Loosemore, R.P.W. & Harley, T.A. 
(2010). Brains and Minds:? On the Usefulness of Localization Data to 
Cognitive Psychology. In SJ Hanson & M. Bunzl (Eds.), Foundational 
Issues in Human Brain Mapping.
Cambridge, MA: MIT Press--. just for the record)

Clearly single cell recording  or even an average  voxel with 14 million 
neurons that are response to say "faces", should not be prima facie 
evidence for a "face area" or "face concept area"  or "face 
category/intension area of the brain".

I am unclear what you mean by "symbol" tho in this context.  Virtual 
concepts seems to be a fine way of thinking of conceptual function, but 
still needs some sort of brain mechanism--no?

I think DL models are likely to provide some more options for what you 
describe as a "virtual concept".   The scale of these systems make them 
difficult to characterize in some principled way.   100s of layers, 100s 
of millions of weights 100s of thousands of hidden units.. this is 
simply a qualitatively different type of representational system then 
what we were dealing with 30 years ago.

One recent paper on concept representation and DL you might find 
interesting:
https://www.frontiersin.org/articles/10.3389/fpsyg.2018.00374/full

Steve Hanson

-- 
Stephen José Hanson
Professor
Director RUBIC (University-Wide)
Department of Psychology (NK)
Cognitive Science Center (NB)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20181118/69013b65/attachment.html>