Re: Strong AI Thesis (No Chinese room, I promise)


Michael Olea wrote:
JGCASEY wrote:

Illusionary contours can be detected in V2 but not in V1.
Which is to be expected as they require integration of
information over a large area.

This turns out to be incorrect. Neurons in V1 can be induced to respond to
illusory contours during tasks that require high resolution.



Given there are approx 200 reciprocal feedback pathways connecting the
30+ visual areas in cortex, including V1 being shown connected to 12
others on the Felleman+vanEssen diagram, it's not surprising a priori
that visual illusions will be a factor of combined activity in *ALL*
visual areas of cortex. When mutual conflicts occur between outputs of
the different areas, as feedback effects ripple through the entire
network.

If we take this one step further and realize that some and/or all of
the 200 feedback pathways will contain BOTH local-field and wide-field
spatial visual information, and also can have BOTH excitatory and
inhibitory effects in the same region, then the possibility for
producing illusions is all the more likely.




This effect
was predicted on theoretical grounds. Then experiments were designed to
detect it. Under some conditions neurons in V1 do in fact respond to
illusory contours. See:

Lee, Mumford (2003). Hierearchical Bayesian Inference in the Visual Cortex.
http://www.dam.brown.edu/people/mumford/Papers/JOSALeeMumford.pdf

-- Michael

.

Relevant Pages

  • Re: What is innate and how can we determine it?
    ... there is something unique about the cortex ... "hardware" the rest of the brain. ... neurons. ... Two electronic circuits may be made up of the same parts ...
    (comp.ai.philosophy)
  • Re: Hawkins ideas on building AIs
    ... Neuroscientist Malcolm Young is mentioned in the Hawkins' review ... that what neurons do is listen to both what the eye, for example, is ... the cat cerebral cortex", Journal of Neuroscience 15: ... feature-binding and population coding in primary visual cortex. ...
    (sci.cognitive)