Well how do you know that the image recognition error is a fault of the ML algorithm (because we can’t capture how organic minds learn, as you are suggesting) a

Well how do you know that the image recognition error is a fault of the ML algorithm (because we can’t capture how organic minds learn, as you are suggesting) and not of the learning sample?