[nolemurs]

> I wish that if they’re gonna downrank “good” pirate content then they should do the same (or worse) for the bad/malicious ones.

I'm pretty confident they'd like to but it's a tricky problem. Based on this article, they're downranking based on number of legit DMCA requests - the good sites will generate a lot of signal on that front. The bad/malicious sites won't.

The malicious/scam sites usually come and go quickly, so it's hard to learn which ones they are - if you figure out one site is a scam, and a new one pops up.

A real solution would require being able to make an automatic judgement about whether a site is bad/malicious based on the content of the site, but that's really hard, and it's really bad when you get false positives, so you have to be super conservative about it.

[GW150914]

A real solution would be to employ competent humans rather than build and rely on incompetent algorithms.

[saalweachter]

Congratulations! You just invented the original Yahoo!.

In all seriousness, curation is one of the most valuable skills a human can provide; I do sometimes wish it was easier to find curated content on the internet, but that isn't really what a search engine does. It is something that eg HN and Reddit provide.

[Rjevski]

As much as I’m against algorithms on closed platforms and walled gardens, I believe algorithms are the way to go on the open web, especially since the amount of content there can be infinite.

[nolemurs]

I'm pretty sure the scale of the problem is large enough to make a manual approach completely and utterly impractical.