I've heard enough of these machine learning buzzword bingos every week when things like this happen. The thing is, I don't care. I just want things like this to never happen again.
This is akin to saying "I've heard enough of these NHTSA reports every week when car accidents happen. The thing is, I don't care. I just want things like this to never happen again."
A 0% error rate is a great aspirational target. But it's also important to acknowledge that a complex system may never be perfect. Insisting otherwise is ignoring reality.
Except that these takedowns aren't "accidents." It's more of a catch-all filter that's designed to proactively censor content that even slightly touches on certain matters.
Near zero (not actual 0%) is definitely achievable and within grasp. We just end up playing funding politics because we don't like the price-tag and liberty-curtailment associated with the value of human life and these things we're trying to solve.
Car accidents? Drunk driving? Murders? Dark-web illegal activities? Domestic violence? Those and many others are all solvable problems that we can get very close to zero instances. But right now we're quibbling and putting measures in place that don't get the numbers anywhere close to zero, they just "incrementally" reduce instances and we all applaud and find those meaningful whilst 99% solutions are ignored.
Then we should shut down videos on the internet entirely, because 500 hours of video per minute is not something anyone can manually moderate. We would all love to live in an utopia where everything works perfectly and there's a solution to every problem, but that's not realistic.
Constantly cracking down on videos that slightly mentions a certain topic goes way too far. Do we really want GFW style censoring to maintain an artificial utopia where harmful content™️ won't exist?
A 0% error rate is a great aspirational target. But it's also important to acknowledge that a complex system may never be perfect. Insisting otherwise is ignoring reality.