The information capacity of human brain

Information capacity of brain Q can be estimated as: Q = N·I, where N is a number of neurons in human brain; i is information capacity of a single neuron.

This capacity can be expressed as possible number of information connections L for each neuron multiplied by possible information capacity of one connection between neurons consisting of address component capacity of the connection and information capacity of logical weight of the connection, a number of distinct states of connection weight W = log2(D). According to estimations based on geometrical measurements of neurons and connections between them, each neuron input can connect neuron with around R = 100,000 outputs of other neurons, which requires around A = log2(R) = 16.5 bits (2 bytes) for addressing. In electronics measurements of voltage and resistance with precision up to 1/5000, i.e. up to 12-bit (1.5 bytes), can be easily accomplished and therefore

Q = N·L· (A + W) = (109 · 15) * (10 000) · (2 + 1.5) ≈ 500 · 1012 bytes = 500 terabytes.

Obviously, a number of possible permutations of the input connections of every neuron with the same set of source neurons should have been taken into account. This easily can be done, but there is no sense in doing it because of (many-fold) scatter in the estimates of the number of neurons and connections between them.

This assessment of human brain information capacity allows, though roughly, but clearly and quickly to assess economic expediency, i.e. effectiveness of various produced devices that strive to be compared with human brain. It is clear that only devices with a comparable information capacity can be compared to the brain.