Best of the best, top of the class is not infallible. It's good enough to sell, because even something as low as 80% accuracy is good enough for things like "Do you want to subscribe to weekly pop tv news", it's not good enough when at the other end you get your account banned.
Google has, on this very site, built up a horrible reputation of using automated processes and having too many users to give those affected by those processes some good recovery. What you're describing is a recipe to replicate that.
99% is not good enough. 99.99% accuracy gets you 0.01% false positives. That's crazy low, right? It's also 100k users when you have 1bn users. Those fancy AI processes are nowhere near that.
I think a lot of their abilities are overstated.
I've met three people in my life who were employed primarily to place ads on the internet. All three were glorified secretarial drones who fell into the position because nobody else wanted to do it.
Any person who knows how to effectively place Internet ads has such high-status skills that they don't have to actually do it.
We^W They may have shitloads of data, but I'm very skeptical we^W they can use this data in any meaningful way. At least, for now.
I don't know what's wrong but despite all that data giant AD companies are supposed to have, I never ever saw any relevant ads, even when I've specifically wanted to see one. Unless I've already bought something - then, sure thing, I'm spammed with more of the same (which is, again absolutely useless - I've already made a purchase).
> We don't need a passport or state-issued ID to determine if someone is a real user.
I don't think we are even capable to come to a consensus what does this mean - to be a "real user".
Am I? What about my alter ego, posting about something I don't feel like publicly associating with (like, porn)? What about a whistle-blowing throwaway account I may create if I learn something fishy? Or what about a "thoughts are my own but went through editorial" corporate representative persona account? And that's just the obvious cases.
I'm increasingly seeing this annoying trend of people hand waving about how a given problem can be solved with data/AI/ML/DL.
- now just imagine the speech was pointed at the bad actors. and that's why this is not as easy as you make it sound.
Maybe some of you are, but people like me read HN too ;-)
We don't need a passport or state-issued ID to determine if someone is a real user. That's the laziest solution any tech company has ever come up with, and it's deliberately lazy.
Facebook and Twitter can already tell if you're a human or not, shit Google lets you just click a button to tell them you're a human and then they determine if they believe you or not. All based on info they already have.
We're the best software engineers the world has ever seen on the cusp of an AI revolution... we don't need your passport. We just don't want to find out the truth, so we make it so hard no one will do it.