Like the brain neuron, it is bi-stated.

fires when input summation going above threshold value

Given input-output pairs, can make itself  learn through changing weights.

Used as a building block in Different brand of current NNs even though not used now