Neurons and Programming
- One of the reasons that I can claim I have a
proto-neuro-cognitive architecture, is that neurons are Turing
complete. (I could even get rid of the proto without blushing.)
- Turing complete means (seriously) anything that can be programmed
can be programmed in a language. So, Java, Python and rule based
systems are all Turing complete.
- It turns out that you can do that with neurons too.
- If you buy that everything that humans do (at least cognitively)
is something that can be programmed, if you have enough neurons,
you can model full cognition. (I do buy that. I even buy it for
sub-cognitively, which seems like a dubious concept.)
- While I'm here I should also note that there are a lot of compuational
models of neurons.
- The perceptron is an integrate and fire model (IF); I do most of
my work with Leaky IF models (LIF), and some with models that
have adaptation.
- There are also more sophisticated models like Hodgkin Huxley models,
and multi compartment models.
