12 ms·
There are plenty of adversarial examples for humans too: http://i.imgur.com/mOTHgnf.jpg http://i.imgur.com/mOTHgnf.jpg
by arnioxux 9y ago
There are plenty of adversarial examples for humans too: http://i.imgur.com/mOTHgnf.jpg http://i.imgur.com/mOTHgnf.jpg
- csomar 9y agoIs it though? The human correctly interpreted the image. The problem is that the image was well, not "real". Human have a limit of figuring out what is real and not real based on experience.
- larsiusprime 9y agoI think the point is that these models are often hyped as being proof that we've reproduced human visual systems, and adversarial examples that humans can still resolve are evidence against that. When the adversarial examples for humans MATCH the adversarial examples for image classifiers, that would be evidence of having reproduced a biological system.
- mannykannot 9y agoFair enough, but these sort of examples are not going to make the problem go away.