So I got into an argument today with a colleague of mine about how much processing power it would take to simulate a human brain and that it should be possible within the next 15 – 20 years.
This argument is based on the assumption that consciousness lies at the atomic level and not at the quantum level or some unknown state below. Based on my VERY ROUGH calculations there are approximately 1.75E+26 atoms in a human brain. (http://education.jlab.org/qa/mathatom_04.html) & (http://en.wikipedia.org/wiki/Brain-to-body_mass_ratio)
Now on the assumption that this number is at least SOMEWHERE close the only remaining unknown that I am left trying to answer is how many logical operations it would require to simulate one Atom and its corresponding interactions.
Given those numbers I could take the current supercomputer ratings of around 1.6 petaflops then extrapolate this number using moore’s law. I figure using this methodology I can come up with a reasonable guess of when we could first simulate a human brain.
I am aware that the declared 1.6 petaflop speed only works with certain types of instructions and won’t necessarily translate to those that you may encounter in such a simulation but I figure in the worst case my guess would be off by a few years.
I know there are allot of very smart people here.
So anyone care to indulge me in my daydream and take a stab at what sort of numbers we are looking at and how they came to them?