Computational models of hearing and vision can build up their own idiosyncratic ‘invariances’ — meaning they respond the same way to stimuli with very different features, according to scientists.
Deep neural networks don’t see the world the way we do
