Hacker News new | ask | show | jobs
by asdfasgasdgasdg 2716 days ago
> I don't know what the answer is, but it seems like we should be able to do a much better job selectively returning results to "wall off" the things we want to separate and avoid accidentally letting through, rather than simplistically blocking words, which is clearly what's happening in some cases.

This is just defense in depth. If you know a query returns 90% CP before filtering, even if your filters are 99% sensitive, you're going to get some CP in the first few pages for that set of terms. So if you identify a query as CP-seeking, then you would rather probably just show nothing at all. Of course, the definition of CP-seeking would have to be tuned, but the ratio of legitimate to CP results would have to be a component of that.

> You don't necessarily have to identify every single image to know that certain terms should not be returning results from sites and pages that have a high probability of being porn

We definitely already do that, and have been doing at least since the late 2000s.

> Anecdotally it seems like all we've done is to edge closer to ruining search for legitimate situations without accomplishing much.

Objectively, this is not true. We have accomplished a great deal in terms of CP suppression, and search is better than it ever has been for the vast majority of legitimate queries. Regardless, I'm sorry that you feel that way.