On human neural network

Last wednesday, I was in Hainburg, Austria on a little conference. One of the lecturers told the same old story about human brain being capable of performing 0.2 mathematical operations per second, while the computer is able to perform billions of operations per second.

This statement is not telling anything, it compares apples to oranges.

Human neural network contains circa 100 billion neurons, each neuron having in average 7000 neural connections to other neurons. This is a massive computation power, nothing nearing current computing power. Intel Core 2 Extreme QX9770 processor can do 59,455 million instructions per second. It is also comparing apples to oranges, but every neuron activation (which has cca 7000 parameters) is quite a complex mathematical operation.

It is very difficult to do simulations of biological neural networks on computers these days, even much more difficult is teaching them.

It is also very difficult to do massive ERP computations, solving of difficult equations in fields like quantum physics. This is the reason, we use computers.

That stupid comparison I see all over again is simply false and misleading.


Written by Juraj Bednár //