(January 29, 2015 at 12:37 am)Surgenator Wrote: Just thinking about the memory space needed for the mapping.
Human brain has ~100 billion neurons with each neuron makes over 1000 connections.
A mouse's brain is ~0.03% smaller in mass than a human brain.
If we assume the neurons between mouse and human are rougly the same, then a mouse will have 30 million neurons and 30 billion connections in total. You would need a 64bit integer to give each neuron a unique id. You will also need 64bit integers for each connection. So that is 64*(30 million + 30 billion) ~ 240 GB of memory just to store the map. This doesn't included other variables (like activation thresholds) for processing.
And as you say that's just for mapping. Some self promoting media scientists have suggested in the past that we'll have computers that surpass human intelligence by 2040 because they extrapolate Moore's law and assume that a single neuron is the equivalent of a single byte.
When in fact a single neuron is an amazingly complicated machine. A good simulation of a real neuron is still far more complicated and capable of far more computation than a whole artificial neural network that uses simple integrators with activation thresholds.
Don't think one byte per neuron, think one core per neuron. And then wonder how you are going to have each core connect to thousands of other cores.
Back to the title of the OP though, here's something to think about. Say your mind was scanned and the data was uploaded into a fantastically large computer and then simulated. How would that computer sense or act? And would you be looking at that computer thinking that you are now uploaded into a machine or would you be aware that you are still standing there looking at it and that you still exist? You certainly wouldn't want to then kill yourself thinking that you were now in the machine.
