> When you are as big as Google, very improbable failures happen all the time, pointing them out doesn't tell the story of how well they're doing unless your expectation is constant perfection
Just a thought experiment, does Youtube need to be as big as it is? Can’t Youtube say “given the current accuracy/precision of ML based solutions and how much our manual operations can scale, we are going to rate limit the content that circulates to match our ability of giving the best judgment on what videos to take down and what videos to keep”. Physical stores have maximum occupancy limits, and with coronavirus some even shown proactive rate limiting to say “hey, we can’t ensure transmission control above a certain density”. Those are corporations too, those also have fiduciary duty.
I think “growth at all costs” is a major driver of these types of failures, especially when costs are externalized and distributed to the userbase; one annoying notification, one abuse feature change, one harmful policy decision at a time.
> Go find rates, not anecdotes, then you might have something
Youtube has near perfect knowledge of the warm data of their operations, and use it to optimize user frustration rather than minimize it, while we only get to know others are being abused when news make it to Hacker News. The inability to find the rates is inherent to the problem. Short of sneaking into Youtube data centers, there is only so much data you can find without Youtube’s say so. Which brings the question, for such a massive policy layer operation, why is this data not made transparent?