4 ms·
That's probably worse, certainly in the short term. Long term, use many uncorrelated solutions: with humans "you can fool all of the people some of the time, a
by ben_w 4y ago
That's probably worse, certainly in the short term.
Long term, use many uncorrelated solutions: with humans "you can fool all of the people some of the time, and some of the people all of the time", but with a single AI (as with all software) once you know how to fool it, you can simultaneously fool every single person in the world who is using it.
- amelius 4y agoOk, but aren't we already using AI in e.g. spam filters? Also, there is some asymmetry in this problem. E.g. a little more false positives (a detection, but no actual threat) is not so bad, so we can err on that side.
- ben_w 4y agoAs I read it, this is the stuff which is already getting past such filters.