[eshvk]

Very cool.

Just wondering out loud reg. your idea to a have a rick roll detector: How exactly would that work? A brute force approach would be to compare how similar the target video is to some sample rick roll videos, I am guessing music matching could be attempted since frame by frame duplicate detection of the video is insanely hard.

From a machine learning perspective, it gets harder because a good rick roll entices the target into clicking the link which turns out to be a false positive. On the other hand, the link could be completely normal so the features of the link itself or not necessarily useful. Hmmm...