Sure, but at the same time, there's the concept of adversarial examples not being the model's fit to noise, but simply being a different, human-imperceivable fe

Sure, but at the same time, there's the concept of adversarial examples not being the model's fit to noise, but simply being a different, human-imperceivable feature instead. Considering that ML has access to orders-of-magnitude greater quantities and resolution of information that we do, why do we expect their decision structure to be intensely linked to our own decision process.