Can cognitive functioning of the human brain change the physical state of the brain? E.g. does self-awareness, self-reflection change the number of neurons or synapses among neurons? I.e. I am trying to develop self-modifying neural networks as described in my other question https://ai.stackexchange.com/questions/7966/dynamic-self-improving-self-modifying-neural-networks-which-can-simulate-goede and I am interested in the biological process that can be analogues of self-modifying neural networks. What determines the physical structure of the brain - number of neurons and synapses and are the cognitive functions among the determinants? I guess that generally the brains can be less developed matter comparing to the artificial neural networks that humans can invent or that can self-evolve in silica after sufficient period of time and that is why human brain can be more restricted in its self-improving capabilities than in silica ANNetworks. Similar situation is with genetic algorithms which can allow unrestricted number of sexes while biology has at most two sexes.

  • 1
    $\begingroup$ Your question has loads of unjustified assumptions. $\endgroup$ Sep 14 '18 at 15:47
  • 2
    $\begingroup$ There might be an answerable question buried in here (the title question is somewhat answerable), but it is obfuscated with a bunch of other conjecture that is unsupported and not helpful towards clarifying your question. If your interest is just in ANN, you should stick to ANN and understand that the connections with biology are analogies. If you want to learn about biological neural networks, you should start with a more basic understanding of neuroscience first. $\endgroup$
    – Bryan Krause
    Sep 14 '18 at 16:01
  • 1
    $\begingroup$ @TomR Bryan is right; interesting though the metaphor is, the more you learn about machine learning and neuroscience, the more you realise the comparison is almost useless. $\endgroup$
    – James
    Sep 17 '18 at 12:23
  • 1
    $\begingroup$ Also agree with @BryanKrause. Here is a link to a related answer of mine, although it deals with the case of a single biological neuron at the phenomenological level. $\endgroup$
    – vkehayas
    Sep 18 '18 at 20:11
  • 1
    $\begingroup$ @TomR Biological networks can be motivation for trying new tricks in ANN, but the cost functions are very different for biology versus silica. Some computationally intensive artificial computations are done effortlessly in biology, and some biologically implausible approaches work great in ANN. We're far enough into understanding both that they have diverged: if you want to make a better ANN, study ANN. If you want to learn about biological neural networks, study biology. $\endgroup$
    – Bryan Krause
    Sep 18 '18 at 20:35