It's easy to think that neurons are essentially binary, given that they fire an action potential if they reach a certain threshold, and otherwise do not fire. This superficial similarity to digital "1's and 0's" belies a wide variety of continuous and non-linear processes that directly influence neuronal processing.
For example, one of the primary mechanisms of information transmission appears to be the rate at which neurons fire - an essentially continuous variable. Similarly, networks of neurons can fire in relative synchrony or in relative disarray; this coherence affects the strength of the signals received by downstream neurons. Finally, inside each and every neuron is a leaky integrator circuit, composed of a variety of ion channels and continuously fluctuating membrane potentials.
7 comments:
Why can't we model this using digital computers? You can model all kinds of "continuous" variables using computers. Why can't you model the complete neuronal stucture of the brain?
They model weather patterns and nuclear reactions and whatever else using computers. No one suggests that the weather works like a digital computer.
Sure, modeling is all well and good (and very, very useful, I think); that's "Weak AI." There are, however, a lot of people out there who are proponents of "Strong AI," contending that--given a high enough level of sophistication and the right programming--a digital computer isn't just a model of a mind, it just is a mind. That's where I think these differences become relevant.
What I am saying is that if you have a sophisticated enough model of the brain, you can recreate the functioning of a brain. You can get to strong AI. The brain isn't hardware it's wetware. But if you model that wet functioning closely enough, and can produce the same responses to stimuli as a real brain, I don't see why you wouldn't have yourself a mind there.
Honestly, I don't think anyone currently interested in strong AI believes that the brain functions like a digital computer.
I'm afraid the specific points you raise are not relevant.
Most versions of multi-valued logic (also modal logic) are in effect equivalent to Binary logic. The only difference is that semantic becomes simpler - it is easier to write complex formulas using, say, 5-valued logic that it is in binary logic. But there is no qualitative jump there. If that were the only difference between a brain and an electronic machine, that the machine could be a perfect model of a brain - just slightly bigger.
You are missing the point. "Strong AI" in the old days meant, let's write algorithms to mimic the algorithms used in the brain, because the brain is like a digital computer. As neuroscience progresses, we learn that the brain is not a computer running algorithms. But it is entirely possible than you can MODEL a brain, down to every last synapse, on a computer. You may be able to create a functional equivalent to a brain, with fine-grained enough modeling of "continuous variables" that the model is functionally equivalent.
The article talks about "the rate at which neurons fire - an essentially continuous variable." If the computer can get lose enough to a "continuous variable" that may be sufficient to functionally model neurons. Even though neurons aren't digital, that doesn't mean you can't create a functional-equivalent model with a digital computer.
Note, I'm not suggesting that if you had a functionally equivalent model of a brain, you'd necessarily have a conscious mind emerging from your desktop. I don't know what would happen. But just because neurons aren't binary doesn't mean you can't model their functioning using a digital computer.
Wow! I've been writing about this topic since I started my blog, but this post seems to have touched off a firestorm of discussion! I'll try to address some of these concerns in a separate post. Thanks for the comments, guys!
Post a Comment