2 ms·

You'd think that google is smart enough to give content a score based on popularity and sheer amount of content (and existence time) and allocate those to human

• bryanrasmussen 5y ago
  perhaps someone having setup the automated system should think, hmm, false positives are rare, but they happen on single items, meaning that people who have lots of items increase their chance of having a single false positive in their collection and thus we will have situations that will seem unfair and idiotic to people outside Google if we remove access to collections of data for a positive hit for violating content unless we put a process in to handle these false positives on large collections of data. But of course this would require the creators of the automated systems for detecting violating content to have a healthy measure of self-doubt regarding the perfection of what they build, which is the kind of thing Google's hiring process seems likely to weed out from what I've read. Thus my conclusion Google is unlikely to ever build an automated system that will take as a central precept that automation is imperfect.
  • Tozen 5y ago
    Very good points. The longer the person has been using YouTube and the more data accumulated, clearly the math shows the more likely they are to get hit with false positives. The flip side of this, is where YouTube/Google is finding excuses to conduct "purges" of data and older accounts. They know what they're doing is wrong, but it has plausible deniability to it. So easy to pull, "Don't blame us, blame the algorithm." As if it isn't people programming and supervising it, and there isn't loads of user complaints as to what's going on.