shipping-isnt-morality:

This isn’t really the important part, but it does annoy me slightly, so let’s get something straight: what’s happening on tumblr right now, with millions of false flags, is par for the course for any A.I. filtration system. It’s not like they got some budget-bin A.I. - this is just about where the technology is right now. I mean, think of those annoying as fuck recapcha things that tumblr made us do a thousand of (foreshadowing?). It works because A.I. in 2018 is still absolute shit at determining if it’s looking at a stoplight, or a car, or a bike, or anything. A.I. is lightyears away from the visual comprehension of a six month old human.

it’s not that they implemented it wrong; it’s that removing NSFW content from a website that allowed it for a fucking decade is an impossible task. A.I. that can reliably distinguish between landscapes and nudes doesn’t exist yet. Text and language recognition is only moderately better, if the users are even speaking any sort of standard English, which I think we can all agree nobody on this website fucking does.

This isn’t really a defense of what’s happening, but just a reminder to stay focused: what’s happening with the false flags isn’t a surprise. This isn’t some wacky glitch; this is what the technology is right now. The best A.I. in the world right now couldn’t do what tumblr’s trying to do, and they knew that going in, because I could have told them that and I’m a person who just occasionally reads the Technology section of Apple news. Maybe they didn’t know it was gonna be this bad, but the best image recognition software in the world can only figure out what a clear well lit photograph is of about 70% of the time, so they knew it wasn’t gonna be good.

They ripped tumblr apart for money. And like, I don’t have a whole lot of positive feelings left about this website, but still: fuck them.