3 ms·
Just manually review top K websites and ban such garbage? Sometimes dumb, bruteforce and biased solution can work way better than any automation you can come u
by nurumaik 2y ago
Just manually review top K websites and ban such garbage?
Sometimes dumb, bruteforce and biased solution can work way better than any automation you can come up with
- dotancohen 2y ago[flagged]
- danielheath 2y agoScale to what? You could manually look at the top million searches each year with four average college graduates.
- carstenhag 2y agoIn all languages of the world? Is that realistic? How could you know whether something in Vietnamese is spam or not? In my opinion it could work, but you'd definitely need 40-80 and not 4 people.
- adgjlsfhk1 2y agoGoogle has ~180k employees. that seems like a pretty small price to pay to make search work...
- disgruntledphd2 2y agoAnd they've been doing user testing of search results for well over a decade (FB started at least 10 years ago, and Google were well into it at that point).
- consteval 2y agoNot everything needs to be computerized and completely automated. You're allowed to just... use human labor. Sure, if every phone call requires an operator that doesn't scale. Okay, so we now have automatic routing. But having a receptionist to route company requests, which may be complex and may not map nicely, still makes sense. It's a human operator, and you pay them. So what, it's a small expense.
- NelsonMinar 2y agoI think that'd be a good approach. There was an idea in that time that everything had to be algorithmic, that hand-ranking or singling out individual sites was a bad idea. Ie: tweak a generic algorithm and then test what it does to top garbage sites, don't just penalize the site directly. I think that's not a bad principle but in practice it didn't seem to work well.