Hacker News Reader

Comment by not2b | Hacker News Reader

They can't; the web is too large to work without it, and it is (legitimate) links that make it possible to distinguish a well-known site from a pop-up plagiarist that just copies all of the text. But for it to work on the modern web you need to be able to distinguish high-quality links from bogus link-farm links.

View on HN

It's more that they seem unwillingly to admit that purely automated methods don't achieve optimal results when they're being fought by equally automated scams. They could bump their overall ranking quality significantly by blacklisting known bad-actors that scrape real sites and put up hollow shells of barely edited, machine-munged plagiarism, yet do not.

And that’s fine. PageRank is/was a genius idea that should still be used today. But they need to do more. When I search something and half the results are Markov Chain like blogspam full of ads, I have to wonder where the effort is going into those things (definitely not AdSense /s).

If anyone can combat it, Google can. It doesn't seem that hard to me to assign low (or negative) weight to links that suddenly appear on a page that hasn't otherwise changed in years, for example. Or links to the same site on pages that otherwise have nothing to do with each other in terms of content/subject. I'm sure it's harder than that, but all those engineers ought to be earning their six figure salaries.

But how do you translate that into (short term) OKRs and work that engineers can build a career upon as an achievement (for promotion)? I think there is a cultural mismatch between how engineers grow and what they work on at Google versus the quality of the search results. There seems to be very limit upside and huge potential downside to working on this at Google.

That said, I don't work for Google and my conjecture is based on the hand wavy details (from engineers that do/have) posted online.

Reply on news.ycombinator.com