Human Brain Computation
I would like to debunk those popular claims that the brain does incredible amounts of computation.
Typical Extreme position:
100 billion cells, each with 1000 synapses to other cells. Speed of the order of 100 Hz. Multiply these to
get 10**(11 + 3 +2) = 10**16 Hz.
"Obviously, no computer will be able to do such things before the very far future:
Criticisms of popular extreme position:
…---The same signals go to each of the 1000 synapses from each cell.
Also, most cells have far fewer functional connections.
…---Most of our human brain-cell connections are organized into local “vertical columns” and the "up"
and "down" connections within the thickness of the cortex are much denser than connections that
spread from side to side. The columns have about 0.5 mm apart and contain the order of 100,000 cells.
Most theorists consider this to suggest that the brain has the order of a million “functional units”. The
sensory systems may have more, but still, might be comparable to the complexity of the GPU in your
computer.
…---The rest of a human brain is not so different from that of lower animals.
Realistic position:
A present-day personal 2-core, 4 gigabyte computer can be seen as operating at 4 Ghz, but with
*unlimited connectivity access” to 4 billion locations. This would allow each of a million ‘simulated
columns” to have, say, 1000 bytes of memory, which leaves 3 gigabytes for representing those columns’
functions and interconnections.
As for all that noise about massive chemical interactions, etc., I suspect this is largely noise, indeed--and
that the overall columnar structure evolved in large part because it helped to "insulate" the functional
behavior of those columns from all those local fluctuations in their micro-environments. Just as modern
computers evolved from analog computers to be composed of flip-flops rather than op-amps!
Other's comments:
EXCEPT for speed, there is no advantage to parallelism. Nothing fundamental is added.
Rubbish. You forgot parallel concurrent interactivity and indeterminacy
Next
I would like to debunk those popular claims that the brain does incredible amounts of computation.
Typical Extreme position:
100 billion cells, each with 1000 synapses to other cells. Speed of the order of 100 Hz. Multiply these to
get 10**(11 + 3 +2) = 10**16 Hz.
"Obviously, no computer will be able to do such things before the very far future:
Criticisms of popular extreme position:
…---The same signals go to each of the 1000 synapses from each cell.
Also, most cells have far fewer functional connections.
…---Most of our human brain-cell connections are organized into local “vertical columns” and the "up"
and "down" connections within the thickness of the cortex are much denser than connections that
spread from side to side. The columns have about 0.5 mm apart and contain the order of 100,000 cells.
Most theorists consider this to suggest that the brain has the order of a million “functional units”. The
sensory systems may have more, but still, might be comparable to the complexity of the GPU in your
computer.
…---The rest of a human brain is not so different from that of lower animals.
Realistic position:
A present-day personal 2-core, 4 gigabyte computer can be seen as operating at 4 Ghz, but with
*unlimited connectivity access” to 4 billion locations. This would allow each of a million ‘simulated
columns” to have, say, 1000 bytes of memory, which leaves 3 gigabytes for representing those columns’
functions and interconnections.
As for all that noise about massive chemical interactions, etc., I suspect this is largely noise, indeed--and
that the overall columnar structure evolved in large part because it helped to "insulate" the functional
behavior of those columns from all those local fluctuations in their micro-environments. Just as modern
computers evolved from analog computers to be composed of flip-flops rather than op-amps!
Other's comments:
EXCEPT for speed, there is no advantage to parallelism. Nothing fundamental is added.
Rubbish. You forgot parallel concurrent interactivity and indeterminacy
Next
 | |  | |  | |  |