RE: Programming the Human Mind:
November 21, 2015 at 6:34 am
(This post was last modified: November 21, 2015 at 7:14 am by I_am_not_mafia.)
(November 20, 2015 at 9:40 pm)bennyboy Wrote: The problem with programming the human mind is that the mind is analog in a very interesting way. Each neuron functions independently in a sense. Yes, there are transmissions of data acrosss synapses. However, there are also (at least as I recall from psych class) free-floating hormones etc. that affect reuptake of neurotransmitters and so on.
In other words, the brain not only consists of independent but cooperative parts, but each part essentially has many billions of parallel processing units.
Even a very advanced computer can't do that.
So I wouldn't worry about the software so much as the hardware. I really don't see how we can reproduce the human mind without essentially reproducing the human brain.
Also each neuron is actually a really complex computational unit. A lot of people think that all it does is integrate its inputs and send out a signal. That misses out on the fact that each neuron has a dendritic tree with compartments with their own gating and effective resistance to the signal travelling up. You can actually perform binary logic on a dendritic tree based on which neurons are connecting to it. There is also suggestion that each dendritic spine on the tree performs a computation as well. Neurons also habituate, internal messengers and a membrane that lets in different ions at different rates that cause internal changes.
For example, what we call the synaptic weight, the strength of the connection between two neurons, is actually made up of three variables; N, P and K. I can't remember which stands for which but this is the number of vesicle release sites, the strength of the signal and the chance of releasing a vesicle.
Each neuron will connect to many orders of magnitude more neurons then the kind of connectivity we see between processors in a computer. Intel still only have a handful of cores and lock contention can really slow down parallel processing. Graphics cards have about 2,000 cores but you they need to work on independent parts of a problem, e.g. different parts of a picture file. Whereas each neuron connects to thousands of other neurons.
You can devote an entire computer to simulating just one neuron. The Bible of Computational Neuroscience is The Biophysics of Computation which describes how a single neuron can compute. Each chapter is devoted to a different part of the neuron, such as the axon. the membrane etc. If you want to skip all that though then just jump to Chapter 14 which provides useful models.
The brain is actually best understood as a complex physical system, not a computational unit.
