Wednesday, Debtember 18, 2024
I'm adjusting my tin hat to talk a bit about Google banning people
Another day, another YouTube channel gets demonitized. While it matters to the channel owner that the channel was demonitized, for this post, it doesn't matter which channel, because it probably happens many times per month. Maybe per day, given the sheer size of YouTube these days. And in every case you do hear about, the owner is going around in circles, with different parts of Google trying to pass the support issue to another part of Google hoping the person with the support issue goes away.
It's not that Google can't fix the issue, it's just that it's not something that is easily done, or cheap. And it's not something any one employee can fix, due to how the whole software gestalt that runs on the computers and handles everything at Google works. And this I don't think is often talked about, or even known. Heck, it's an educated guess on my part, but with Google big on automation (because as they have stated often times, “humans don't scale”) they just can't (or won't) manually deal with the 500 hours of video uploaded every minute at Google (as of June 2022, which seems to be most quoted value I found with minimal searching).
I suspect that Google engineers have written a massive machine learning system with so many variables that it's impossible to figure out why any one person has been demonitized, banned, or removed from the Google system. And I don't think it's ever just one thing that caused the Google AI to “pull the trigger” so-to-speak, just many little things that happen to converge on a particularly day to cause the Google AI to go “Thou shalt be penalized!”
I think that explains why Google employees are cagey about the cause of a ban. They explain the lack of explanation with keeping reasons from Really Bad Actors™ who would learn to walk the line but not cross it. It's an excuse—a convenient one, a plausible one, but one that hides, in my opinion, the truth—the Google employees don't know the reason! It's an opaque wall of AI software. It's like asking you “where is your name stored in your memory?”
And it also explains why it takes a massive PR campaign on the part of the punished party to force Google employees to rectify the situation—that means running the taining set, again, with yet another exception to the banning rules. That takes time, and energy, and is probably something Google's upper management don't like doing that often.
Also, not to sound like a Google apologist, but I do wonder how many Bad People™ Google actually ban versus the false positives you hear about in the news. Given the scale, it's could be a rather large number of Bad People™ are removed from Google all the time. But it doesn't seem easy to come across such figures. Hmmmmm …
