Research on artificial neural networks comes from the observation that human and animal brain operates in a very different way from a digital super computer. It can perform tasks which would result very difficult or impossible for a computer.
The brain is characterized by a very large number of neurons (up to 10 billion in the cortex) with massive interconnections (60 trillions) and by a relatively slow rate of operation (in the millisecond range). In simple words it is a complex, nonlinear, and parallel computer.
With similar structure, an artificial neural network is a massevely parallel ditributed processor, which can store and retrieve experiential knowledge; knowledge itself is acquired through a learning process, and stored in interneuron connection strengths (known as synaptic weights).