News for hackers. Stuff that matters.
Re:Sometimes, requiring explanability is counterproductive by esyir on Thursday February 06, 2020 @01:27AM (#22253874)
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.

There are no comments on this comment yet.