Page 62 - Vidyalaya Magazine- Kendriya Vidyalaya Rishikesh

The experts estimate that if we compare our brain to a computer we can store approxi-
mately from a couple of terabytes to approximately 2.5 petabytes (~2,500,000
GB !!). However, this comparison is limited by the fact that the way the human brain
creates and stores memories is nothing like a computer.

The brain has about a billion neurons and each neuron is connected to about a thousand
other neurons which makes about a trillion connections. That is a rough estimate of the
hard wiring. The way the neural networks combine allows a single neuron to be involved
in many memories, which takes an estimate along an exponential curve up into the tril-
lions.

The possible number of unique combinations of inputs for a single neuron with just 100
incoming dendrites could be computed as 100 x 99 x 98 x 97 x .... x 2 x 1 possibilities.
That represents more than 1, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000,
000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000,
000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000, 000,
000, 000, 000, 000, 000, 000, 000 unique possible combinations! Multiply that number
by 100 and divide by 8 to measure the number of bytes of possible memory. A single
nerve cell with 100 dendrites can potentially remember that many bytes of singular
combinations. Some nerve cells have up to 2,50,000 dendrites! Only the possible exist-
ence of such codes can explain the phenomenal capacity of human memory.

Brains store information through a complex physiological response to chemical stimuli at
the synapse. Bitwise computer memory and internodal synaptic "memory" are fundamen-
tally different because synapses don't have a "state" or a "switch" that they can use to
define a bit of information (binary encoding)...they encode this information as a synap-
tic weight or connective strength that can change as the network learns more about the
stimuli.

While neural networks can perform extremely complex computation, this is only possible
because of scale and plasticity. A neuron on its own is more of a filter than a traditional
computing unit--it transforms input in a controlled fashion into a novel output.

Prateek Gupta
Computer Instructor
KV Rishikesh