The learning process in artificial neural networks involves connecting currently active cells, instead of adjusting all the weights, using a simple process called backpropagation to adjust the neural connections.