More on Modeling the Human Brain with Computers

I was pleased (humbled?) to get a polite and not-angry response from Timothy B. Lee in regards to his dissent that we can model the human brain via computers. He does beg me for a reply:

I agree with you that an insect brain would be a good starting place. Do you know if there are any efforts in this area? I would be surprised if anyone has succeeded in simulating an insect brain in software, but if that happens it will cause me to re-consider my argument.

Here, you can find an IBM press release about trying to build a cat-sized brain. Here, you can find my exact argument (written by someone who actually knows this stuff) that a human-brain model is such a monumental task that scientists should, and are, building bug models first.


But I think Lee, in his well-written reply, overstates the peril in modeling a complex system:

I'm not a neurologist, but as I understand it, neurons do all sorts of interesting things that aren't really captured by the models you described: they grow new connections to other neurons, they change their behavior in reaction to changes in the concentrations of various hormones, they're tightly coupled to various other bodily systems like your eyes and your muscles, and so forth.

The question is whether you could build a software model that modeled all of these relevant behaviors while still being reasonably efficient. I'm skeptical on both counts: I doubt we'll ever have a mathematically tractable model for every relevant aspect of a neuron's behavior (we're still struggling with protein folding, for example), and I think that if we did come up with such a model, it's likely to be sufficiently bloated that we won't have enough computing power to run 100 billion copies in parallel for many decades.

I think he has a valid point; exactly mimicking a neuron in every way is probably impossible, either with hardware, or with software as I suggested. However, I think we need to ponder the following: can nature be computationally modeled accurately using simplifications?
By and large, the answer is yes. He admits it, really, in his original post. If we simplify the weather and model it,we can fairly accurately forecast it...at least in the short term. And the thing is, if I want to know the weather forecast for 24 hours in the future I don't need all the world's supercomputers combined doing a massive CFD simulation. I just need a single weather station's Doppler and a decent software program.
Another example would be computational dynamics modeling of human-made structures, like bridges and buildings. Sure, a fair argument could be made that the only truly accurate model of a bridge would be one that included every single iota of concrete, perfectly modeled the wind blowing past it, and perfectly modeled each and every possible combination of cars and trucks that could ever go over it (including perfect models of those vehicles). But for the sake of sanity, a simplified computational model with a few million nodes can give a pretty damn accurate picture of just what might happen to the bridge in the next 25 years.

I'd be an arse if I tried to say dynamic modeling of a bridge was anywhere near the complexity required to model the brain. But nevertheless, what I am getting at is that increasingly complex models of neuronal interaction and increasingly accurate models of the behavior of a neuron could create more and more realistic models of a human (or insect or cat or whatever) brain. Eventually, its not impossible to believe that we could hack together some sort of simulation that actually was accurate enough to mimic a brain.

I think people who suggest the Technological Singularity will occur in their lifetime are the same as people who suggest Jesus will return during theirs or people who expect to make contact with aliens during their lifetime. I do not know if it is arrogance that makes a human think that the influential events will occur while they are alive (not after or worse - before), but something about us makes us want to "be here" for these events. I do not agree with Tim Lee; the day will come when humanity constructs a computational simulation of a human brain that equals our biological one. The day will come with biology and technology blur and meld. But I find it extremely dubious when people suggest it will happen by next Tuesday.


_