Neural networks: inspiration from the brain

The brain contains billions of nerve cells (neurons). Each cell receives signals from other cells. If the combined signals are strong enough, the cell sends a signal onward.

A neural network in a computer works on the same principle: small units receive numbers, sum them up, and pass on a value if the sum exceeds a certain threshold.

Hence the name. But it is an inspiration, not a copy.

Here is how they differ:

	Brain	Neural network in a computer
Number of units	~86 billion	millions to billions of «weights»
Learning	constantly, from little data	in a training phase, from enormous data
Energy	~20 watts (a light bulb)	often thousands of watts
Understands?	yes	no — it calculates
Can feel?	yes	no

A model that generates text about emotions feels nothing. It has learned which words statistically tend to follow one another when people write about emotions.

Keeping «inspired by» and «same as» separate is the most important point.