[exporectomy]

A lot of comments here seem to be overlooking the point of the article. That brains predict their sensory inputs isn't new. What seems to be is that a NN where correct predictions are done with small values of the weights and error correction has large weights uses lower weight values overall. That being energy efficient seems to be more a coincidence of how biological brains happen to work - larger weights use more energy. At least that's how I understand it. Doesn't seem quite as incredible as predicting sensory inputs.

Obviously being able to predict the future has value beyond conserving energy.