Exploring the Space of Adversarial Images — mentioned by cnewey

Tue, 12 Dec 2017 14:28:55 GMT

You're implicitly making the assumption that an algorithm is unbiased and less likely to be fallible - and unlikely to be susceptible to bias or manipulation (at least, compared to a human). There is a considerable amount of academic and popular literature showing that these assumptions are false, at least at a technical level. The problem of vetting YouTube videos isn't necessarily something that can be solved by an algorithm, at the end of the day. Lots of the simple cases, sure - but such an algorithm won't work on even the most moderately challenging content. After all, moderating complex…

Paper and discussion history. A mention is not an endorsement.

Exploring the Space of Adversarial Images — mentioned by mholt

Sat, 15 Apr 2017 16:22:39 GMT

This is called "adversarial training" and is ineffective because of transferability. Also the space of adversarial examples is thought to be large (especially as dimensionality increases), making robust adversarial training intractable in practice. See https://arxiv.org/abs/1312.6199 and https://arxiv.org/abs/1510.05328

Paper and discussion history. A mention is not an endorsement.

Exploring the Space of Adversarial Images · HackerPapers discussions

Exploring the Space of Adversarial Images — mentioned by cnewey

Exploring the Space of Adversarial Images — mentioned by mholt