Your Mind May Not Even Be a Biological Neural Net

My Mind Is Just A Neural Net

The important thing to remember about a biological neural network, and one that is often overlooked, is that it is a theoretical functional and possibly structural component of a brain, specifically a human and some non human animals brains. In my view it is something of a stretch to call it a theoretical component given the relative paucity of hard data in support of it, however, I will grant it is stronger than a hypothetical function/structure at least. What has been described in the above linked article is an artificial neural network from which the author has gone on to philosophize about various aspects of the mind and mind/body problem. The cart has been put way before the horse as it were as no where has it been established that an artificial neural network is anything at all like a mind. We do not even know what a mind is like (at least in so far as the existence and structure/function of hypothesized/theorized neural networks). To suggest that by mimicking/aping/copying/taking as a model a biological neural network (which remember is a theoretical/hypothetical thing and may not be an accurate representation at all, and could in fact be completely wrong) with an artificial version, we can replicate or have replicated intelligence or learning or any aspect of a human or non human animal mind is absurd. It is another aspect of the compulogical fallacy and it is patently ridiculous on its face.

If you want to tell me your machine is intelligent or ‘learning’ (another term for which disagreement as to what it actually is or how it happens in a human person abounds) that is fine. I will disagree and I will try and prove you wrong, but I will believe it is possible. However, if you tell me you have replicated or mimicked a human brain to make your machine intelligent or to make it learn I will tell you it is absurd and (for the moment at least) impossible and not give you the time of day. I have said this before but I guess it is time to say it again, it is far past time for the parting of the ways of computers/technology and biology. Stop focusing on the human brain/mind and trying to prove how computers are ‘like it’ or not. Leave biology to the biologists and take care of your own kind. Biologists are not wasting time and energy trying to make us believe in ‘living’ machines (at least not any of the sane variety) so why are computer scientists/technologists always trying to make us believe in ‘learning’ or ‘intelligent’ ones?

You might object with the classic canard "but all scientific principles are simply theories" If I can say a biological neural network is just a theory I can also say gravity is just a theory. Or evolution by natural selection is just a theory. In part you would be correct to say this. Outside of the universal laws (gravity is most probably one of these) all science is theory. It can always be proven to be incorrect or inaccurate over time as additional evidence develops which contradicts it. That said, nobody is currently attempting to use this theory as the basis for developing computer hardware/software that mimics it either, neither are they attempting to use the theory to develop computer hardware/software with human level intelligence. The same can be said for the "theory" of gravity. It is one thing to say that all science is theory and quite another to actually move about and live in the real world as if that is the case. We simply accept these theories as the best we have and continue to refine them. In the meanwhile we "act" as if they are correct. If we did not we would be totally frozen, unable to progress. One cannot wait for the final "proof" (and there is actually no such thing) of any theory before predicating future actions (e.g. designing future experiments) on the belief in its correctness.

Why not then just "believe" that biological neural network theory is the correct explanation for human intelligence and go ahead and design our artificial neural networks based on that theory, as we have done? What is the problem here? The problem lies not in the actions taken, but rather in the explanations and beliefs in the outcomes of these actions. Artificial neural networks designed based on the principles of the theory of neural networks work very well. But, they work very well for computing. Because they are part of a computer. They are part of a machine. That is what they were designed to do. ANNs do not work very well, in fact they do not work at all for thinking, or for having consciousness, or for believing in things, or understanding things, or having knowledge of the external world. Many of these things are hard requirements of intelligence and all may be requirements of human level intelligence. ANN's have none of them. What they do have is an extraordinary ability to compute. Thus it makes sense why they are so popular as the basis of many modern computers. Just like machine learning and AI, ANNs are a form of modern computing. Powerful, useful, interesting, better than past computers absolutely. Intelligent, learning, or capable of either of those things, absolutely not.

Author’s note: In an earlier version of this article I took a much stronger position against the theory of biological neural networks and suggested that it was really only a hypothetical functional/structural component of a human and some non-human animals brains. I have partially withdrawn that objection for this version.