(April 17, 2012 at 3:00 am)Perhaps Wrote: 4. Do colors exist?
"Wavelength" exists (for a specific ontology) as the distance between wavecrests in the electric or magnetic field of a light ray. Out sensory apparatus also exists (in a different ontology), and our vision would appear to have evolved so as to differentiate certain wavelengths in a range from other wavelengths in that range. That differentiation is called "color". So it would appear to happily exist so long as you accept that all wavelengths in the visible spectrum do not stimulate the various eye cells identically (and for deeper reasons to do with subsequent operations in the visual signal chain).
It does not matter that color is absolutely the same between individuals, because it is a relative value - it exist to show differences. If what I see as red, you see as blue, so long as we both call it "red" (a process that begins as a labelling when we acquire language by rote learning), then we can meaningfully communicate the color of (say) apples.
Two artificial neural networks, trained on different measurements of the same phenomenon, will most probably have a different set of internal weights on their links. However, when presented with new data, they will produce similar results, providing the training was valid (no overtraining, no undertraining, suitable annealing...). The networks capture the statistical trends in the input data and it is the (differential) relationship between the weights, and the network's structure (connectivity, bias functions and structure) that is of importance, not the absolute values of the weights themselves. So we can see from simple computer models that internal state (analogous to a subjective perception of color) is not the relevant variable, it is only meaningful relative to an architecture: subjectivity is irrelevant.
