Web Brain Idea
I’m studying neuroscience on three different Coursera classes. Its something I always wanted to learn about and they are actually pretty good classes. Learning about the chemical and electrical action of neurons and synapses it occurred to me that a web based simulation of the brain might be possible. This is what I am thinking.
Build a model of each type of neuron – There are several good ones available written in MatLab now and that would be the place to start. The models have to be able to solve differential equations but that is within reach. Make the models open source and have them ported to Window, Linux, Arduino, Android or any other language or computing platform someone wants to implement
Add to the model the ability to communicate with other models over the net. This would replicate the way a neuron receives input from another neuron through its dendrites and how it sends it output on to the next neuron through its axon terminals. Each neuron dendrite may have up to 10,000 inputs and a smaller number of axon outputs.
The neuron simulators would be distributed around the world using the same techniques that are used to push data and software out to the network for SETI@Home.
Most of the simulators would simply stay on line, receive inputs, compute outputs, and pass the result along to the next neuron. They would simulate the response times of a biological neuron.
Some of the simulators would be set to receive as input stimulus from the real world if the owner of the computer wished to configure his simulator for that. Inputs could be from things like digital weather stations (you can buy one ready to go with a capable processor for about $50) or digitized audio or switch closures or anything else with a rather low bandwidth.
Some of the simulators would be set to output to the real world connected to digital displays or banks of LEDs or an alarm horn or a relay closure or any other low bandwidth output.
After connection the neurons would be subject to ‘body wide’ inputs such change of ambient temperature, chemical mixtures etc. The simulators would be capable of reporting back to the master station when pinged for statistics such as number of spikes sent and to where, spacing between spikes, number of inputs to its dendrites etc.
The simulators would not be automatically connected when ‘born’ but would receive instructions to connect to other simulators from a master database. There would be no attempt to solve any problem by attempting to wire the network a specific way. It would be random. The network should ‘prune’ itself when the neuron simulators change there operation due to increasing or decreasing inputs from other neurons just as they do in a biological brain. A neuron would receive no feedback from its target neurons just as it is in a biologic neuron. The network may and produce ‘something’ eventually.
I would never think this is possible simply because of the number of simulators that would have to be deployed – except for one thing. The classes I am taking are attended by an average of 35,000 students all over the world and some have upwards of 100,000 and it stands to reason that that kind of interest could be tapped into.
I’m looking into the ‘Blue Brain’ project in Europe and the Brain Research project announced by Obama but these are several orders more complex than what I’m thinking about. I have no illusion that this would be anything other than a learning tool but it would be great fun.
So what do you think? Have you heard of any other ideas like this? If someone has an ongoing project like it I would really be interested in knowing about it.