This is my provisional position about aesthetics: aesthetics is a two-place word (“X likes Y”), but for human Xi’s, “X1 likes Y”, “X2 likes Y”, “X3 likes Y” etc. are correlated with one another. Therefore, one could draw a network like Network 1 in “Neural Categories” with the nodes labelled “X1 likes Y”, “X2 likes Y”, “X3 likes Y” etc.; but such a network would be infeasible to compute, so one can approximate it with a network like Network 2 with the central node labelled “Y is beautiful”. This is usually useful, but breaks down outside the domain of applicability of the approximation, i.e. when considering stuff that lots of people like and lots of people hate such as Justin Bieber’s music; but even then, a smaller Network 2-type network with only aesthetic judgements of a certain group of people (e.g. musicians, or people like lukeprog who’ve heard lots of different music, or people with IQ above 130, or whatever) may (or may not) be useful.
This is my provisional position about aesthetics: aesthetics is a two-place word (“X likes Y”), but for human Xi’s, “X1 likes Y”, “X2 likes Y”, “X3 likes Y” etc. are correlated with one another. Therefore, one could draw a network like Network 1 in “Neural Categories” with the nodes labelled “X1 likes Y”, “X2 likes Y”, “X3 likes Y” etc.; but such a network would be infeasible to compute, so one can approximate it with a network like Network 2 with the central node labelled “Y is beautiful”. This is usually useful, but breaks down outside the domain of applicability of the approximation, i.e. when considering stuff that lots of people like and lots of people hate such as Justin Bieber’s music; but even then, a smaller Network 2-type network with only aesthetic judgements of a certain group of people (e.g. musicians, or people like lukeprog who’ve heard lots of different music, or people with IQ above 130, or whatever) may (or may not) be useful.