So there's a private option that's terrible (WebMD) and a private option that's terrific (Merck Manuals). And it's the terrible one that ranks well at Google.
Sounds like a Google problem and apparently only a Google problem.
Sounds like a Google problem and apparently only a Google problem.
Filtering out, say, Stack Overflow or Reddit because it has ads doesn't help you when it answers your question and is perhaps the only thing on the internet that truly does.
People seem to think there's this ad-less replica of the internet, sitting right behind our ad-riddled internet, where everything they want exists for free, it's just hidden. In reality, the websites making money are the ones providing the vast majority of things people are looking for.
Use https://search.marginalia.nu if you want to severely punish ads.
Does this site (in general, not just this page) have more than 5 advertisements per page? Between 2 and 4?
Does this site attempt to load 12 trackers? "Only" 4 trackers? Just 1?
Does an AI text analysis of the first few paragraphs match on this nonsense?:
> Fixing your gadget is important. Many people find that their gadget sometimes breaks. Gadget helps us do action easier, and improves our lives. We all hate it when our Gadget doesn't work the way we expect it to. It can be frustrating. Read below for tips on how to fix your gadget. (Followed by 3 more paragraphs of filler before getting to regurgitated gems like "reboot it".
I'm sure we have the AI tech now to semantically see this bullshit and downrank it. Right? (Ok, maybe I overestimate how easy this would be. Forgive me, I'm just ranting here)
That's how the world was before Google.
They already look at things like clickthroughand dwell time and bounce back. If enough people dislike Example.com enough to avoid clicking on it or come back to search after visiting it, the engine learns that it is a bad result.
Maybe the problem is that most people like what you don't like.
This is a game as old as search engines. In 2005, it meant filtering out sites that were just lists of keywords, not coherent sentences and paragraphs. It meant for giving extra points to articles with structure, such as header tags and paragraphs, as opposed to just blobs of text. It meant using PageRank to organically discover which pages real people thought were useful.
It's a much subtler and more difficult problem in 2022, but there are also better tools to do it (big NLG models). It just seems that Google lost interest in quality control at some point.
And I would guess they lost interest in quality control because of Chrome's market penetration. Chrome is a browser monopoly at this point, and with Google being the default search engine on Chrome, they no longer have to give quality results to maintain their search user base. On top of that, they control such a large share of the ad market that any SEO spam website is more likely than not to be using AdSense. Which means they have a financial incentive to deliver page views to SEO spam sites, which tend to have higher ad/content ratios.
See: clickbait YouTube videos that show you something you really want to see in the thumbnail, then spend 10 minutes doing something else before showing it, and when you see it it’s a tiny aside with no more context than what you got in the thumbnail. If it’s even in the video at all.
Those videos have both high clickthrough (thus “click bait”) and also high dwell time (from people waiting for the thing they wanted to see to show up.) They do also have high bounceback, but only from people who recognize what’s going on. “A new sucker’s born every minute”, and those suckers will click the video and watch it, because they don’t yet know the principle that this specific kind of enticing thumbnail+title format implies that they won’t find what they want here.
These metrics all measure, effectively, “wanting” rather than “having.” It’s like measuring food by how addictive it is, rather than by how satisfied it makes you. You’ll end up optimizing toward cheetos — literally flavoured air — rather than toward anything that fills your stomach. People might enjoy cheetos while they’re eating them, but if they’re genuinely hungry, cheetos won’t solve their problem — they’ll still be hungry afterward.
1. it's orthogonal to relevance of content (semi-solvable algorithmically I suspect) 2. it's antithetical to Google's core business model (a lot tougher nut to crack)
I disagree. The way content is presented matters. Splitting an article into 4-6 pages and filling those pages with ads makes me not want to read that content. I'd much rather go somewhere that has the same text in a single page and only a few ads.
The entire point of my comment was that it's not orthogonal. The ads are what fuels the click bait and SEO-driven articles. Nobody for example would ever pay a subscription to a website that is just waffle filler. While stackoerflow has ads, it's much better in that regard to the SEO spam pages.
I hate ads, but I don't think we should be focusing on them here. Some sites that have ads have garbage content, and some sites that have ads have useful content. Just... find the useful content, and return it in search results. I know "just" is doing an awful lot of heavy lifting there, but I don't think "has/does not have ads" is as important a signal to a search engine's algorithm as you think it is.
I want something like webrings to become A Thing again. A user curated search engine. And the users doing the curated need to be vetted. I don't know if this is even possible, but I get tired of having to come to HN to get a human recommendation that is miles better than the algorithmic crap from the current search engines.
It comes back to the age-old question: Who vets the vetters?
You can make decent attempts, such as academic peer review. Even this system perpetuates its own problems (beta amyloid) and has perverse incentives (publish or perish, falsified results), though.
Semantic web had some good ideas about this. Networks of signed FOAF data attached to articles and posts. You could form a side graph of trust information that you could revoke at any time.
Post your bookmarks, share to your friends, encourage your friends to do the same. Import those bookmarks into a search engine site fliter extension.
Search "medical information awesome list"
https://github.com/NeovaHealth/awesome-health
https://github.com/lalaithan/awesome-health
https://github.com/jeromecc/awesome-health
Fork and make your own!
The best algorithms adapt to feedback. Surely Google's own algorithm can accept Google's feedback to adjust for flaws in it.
> For pages about clear YMYL topics, we have very high Page Quality rating standards because low quality pages could potentially negatively impact a person’s health, financial stability, or safety, or the welfare or well-being of society.
https://static.googleusercontent.com/media/guidelines.raterh...
They implement it as agnostic tuning as much as possible, avoiding single human chery picking sites. They use panels of humans (mturk style) for quality ratings.
Could you imagine the outrage if Google said "The government is always the best source about everything?"
What even would be the point? Use the government search engine for that use case.
For example, somebody could search for "best rewards credit card for married couple with normal lifestyle" and find some listicle full of referral links that's a few years old. But the best advice might be in a Reddit or Twitter discussion titled "what's a good card for my P2 with no AF, low MSR, and at least 2CPP" whose replies are terse comments like "CSP?" or "BofA Custom; select CNP transactions as the 3% category and order everything online. they give high CL if you want to AU instead". There's enough jargon and levels of understanding that Google can't find the best advice, only good keyword matches.
I recently had a similar issue when I searched for something well-phrased and generic ("how to stop wood joints from squeaking"). The results were lackluster, but after a few attempts, I found the most helpful results were actually under a specific application ("how to stop a bed frame from squeaking").
So what's my point? It's not Google's fault. They are trying their best of optimize text search, with some fancy word associations and other stuff to help. But it's going to take a lot of effort to make an efficient, scalable, general-purpose AI that can achieve near-human understanding of text and then find the most relevant articles related to that. This "super-Google" would have to comprehend every post and comment on the internet, contextualize the knowledge ("a squeaky bed is caused by squeaky wood joints or fasteners, so this advice is useful for any kind of squeaking wood furniture"), and quickly generate results for every query.
It's not WebMD's fault they have ad revenue that lets them hire writers to SEO their articles with the best keywords. Nor is it Merck's fault that they are using specific language that doesn't cover all the phrases that a person could search. Nor is it Google's for making a search engine without human comprehension. It's just a technological gap that can't be bridged in the present day.
If worst comes to worst, just add "reddit" to your search term, and then all you have to do is determine whether he answers you find look like they came from a human, a spammer, or a corporation.