You can calculate a theoretical maximum speed:
N_i = 100 * 10^9 neurons in the brain (1)
i_hz = 3.0 average spike speed about 3-4.5 hz (2)
S_i = 7 * 10^3 Average number of synapses (3)
Computations = N_i * i_hz * S_i = 2.1 * 10 ^15
But for this number i made the assumption that every spike is pertaining in the calculation.
Which is false. The brain uses a form of population coding because of the massive amounts of 'noise' present in the brain. Both caused by input and physics. If i had to make a guesstimation I would divide this number trough 1000 to get to the basic operating system computations of the brain.
It is possible to then guesstimate the actual 'user' program load by taking the average spike speed of neurons when performing actual tasks (40 hz. , gamma waves) originating from the thalamus. (4)
this gamma wave is the most promising hypothesis for the binding/consciousness problem.
Combine this 40hz with the number of neurons pertaining in the actual active though and you get the user work load.
(you could calculate that from fmri data (measures energy usage) to actual neuron energy usage)
You then have to take in the population coding and synapse number etc. etc.
basically all number you see may be of by a factor of 10^3 or even more
Ow and a pictorial answer:
The image is from kurzweil great read:
http://singularity.com/
Ah.. and if you wonder... I study AI with a focus on neuron computation....
(1)
http://hypertextbook.com/facts/2002/Ani ... iye2.shtml
(2)
http://books.google.nl/books?id=ffw6aBE ... q=&f=false
(3)
http://en.wikipedia.org/wiki/Neuron
(4)
http://en.wikipedia.org/wiki/Gamma_wave