the human process 11 trillion bits per second

bobsmith76

I'm trying to figure out how many bits of information the human has stored in his mind. I personally believe that information is not stored in atoms in the person's brain but is stored in an immaterial realm and the information is not made of material. We can talk about whether or not that's true in a philosophy forum if you want. Nick Herbert in the elemental mind
http://www.amazon.com/Elemental-Mind...4709212&sr=8-1
calculates that we process 11 trillion bits per second. I think this is a low ball number because he thinks our eyes are processing 125 million bits per second, but that's not true because we have 125 million rods/cones and each one is being hit by several photons per second and moreover we can distinguish 8 million different colors so you would need an I don't know how many bit system to distinguish 8 million different colors.

Whatever the number is I think what can definitely be answered here is the ratio of the size computer processor one has to how much information it can process per second. For example my computer has a 160 gig harddrive and it can download 3 megs per second, that's (1.6 * 10^11)/(3*10^6) which is about 10^5.

I'm thinking that if we can get a rough idea of what the ratio of processor size to processing per second is we can get a small idea of the magnitude of the problem. Of course things with the mind might be entirely different.

So my question here is just how much larger do computers have to be if they are to process x bits per second?

My general thesis will be that if we can show that a human has more than about 10^17 bits in their memory then that would prove that information is not material. We only have 4 * 10^10 neurons and at most 5 * 10^14 synapses so it's hard to justify where this information would be located in the brain.

256bits

This sounds surprisingly close to the 16,777,216 different colors that a PC monitor and video card can display at the setting of True Color.

Perhaps your computer can distinguish color number 234,568 from 234,569 but I cannot.
My eyes are just not that sensitive and my brain processing capabilities fail there also.

At the High Color setting the output is 65,536 different colors, and most people do not notice any difference in a picture displayed at High Color from True Color.

But truth be told, I can see an unlimited number of colors far greater than 8 million.

Be a little wary when trying to digitize the brain function in the same way that a typical computer processes information.

Negatron

For the sake of being able to answer your question I will have to pretend you did not say that.

If we assume that the brain is an analog system then there is no absolute answer. We can suggest to say how much data we need to process to replicate the phenomenon to high accuracy. I'll assume an analog nature due to the modulatory behavior of the synaptic cleft, however because the space is so small there is likely a quantization of the degree of resistivity (small number of molecules), in this sense it may be true that there could be a reasonable 1:1 numerical correspondence, although other contributing factors to synaptic potential such as the size of the synapse and it's location would likely suggest that the representation of it's weight should be continuous.

If I represent the 100 Trillion synapses with 4 Byte floats, that's 400 Terabytes of data. If I assume that < 10% of the synapses need to be updated within one second that's 40TB or 320 trillion bits. In this regard I'm just about within an order of magnitude in the estimation.

The problem is however that I don't know

#1. How rapidly synaptic weights can be modified in a biological brain.

#2. The number of time-steps per second to achieve accurate integration. The number of calculations that need to be undergone to update a single synapse would likely be far larger than what the synapse contains due to the numerical simulation required to acquire the modulatory behavior of the synapse. If the temporal model of the brain can be simplified to some sort of discretized representations such as in artificial neural networks then this becomes much easier to analyze, however at this point in time it is unknown what are the ultimate functional relevancies of the working brain.

If I were then to assume something on the order of 100 integration steps per second I would come to a rough conclusion that on the upper end of the estimate about 40 Petaflops of processing power would be required.


Good luck with your thesis. I suspect you will have a very difficult task ahead of you.

Optima

Dear Bobsmith76 I found this article which might help to answer your questions.
h t t p : // ergo. human. cornell. edu/ studentdownloads/ DEA3250pdfs / hip. pdf

Best of lucks,

jdev

I would recommend reading the book "The Singularity is Near" by Ray Kurzweil if he is interested in the topic of the capacities of the human brain and the comparison to computers.

Optima

Agree, the singularity is near is highly recommended.
bests.