Obviously google was basically already unusable over the past five years but this is... this is something else. Any subject you want to look up will have the first three pages entirely comprised of procedurally generated fake blogs and websites. And there's nothing there. Just scraped text from articles mashed together with varying amounts of hallucinations on top. Not even selling a product, just to harvest ad revenue.
What's maybe even worse is the steady degradation of image searches too. Joe Everyman generates some slop, the image's metadata tags include a historical artist's name. Some one on pinterest pins it while trawling the web. Now there are fifty gooey generated sludge images when you look up the historical painter.
And its not just artists. Historical figures, animals too. Look up 'baby peacock' if you want a clear example of it.
It's a pretty funny bit google, I hope it keeps going.
Pinterest has ruined google image search. Now everything is pins that are blurry compressed images. I have to use yandex's reverse image search to find the source. Thank you Russia.
Yandex's reverse image search is so much better than anything else it's actually surprising to see a technology that is that much far ahead of others.
Russia is such a primitive and backwards place. They are decades behind the west in developing advanced enshittification technology.
I did my part by "teaching" a chatbot by just feeding it answers from some other chatbot. I never confirmed if anything was correct myself.
:im-doing-my-part:
When someone says "you blew it up! you blew it all up! you maniacs!!!" they are angry because they think someone else made things so bad that everything fell apart. It's like they're saying "you kept blowing that balloon even though it was about to pop, and now look what you've done!"
Earlier today I was trying to remember the name of the game Supraland, and typed into Google "video game where you play as a toy in a back yard." It was really good at this sort of vagueness-based question back in ~2014 when they switched to their "knowledge graph" tech. Now it shows reaction videos to the most recent game to fit that description, inexplicable results from the US patent office, and then endless Amazon results for physical lawn toys.
Type anything into Google and you will get 100 identical articles "Top ten insert subject"
Literally all the same just copy pasted
Even the major news outlets are like this
I'm pretty sure the only information google will still reliably point you towards is where to buy things :agony-minion:
Almost every non trivial information search forces me to add "site:reddit.com" to the query to only get Reddit results. That is the only way to get rid of the stupid fucking top 10 lists and other regurgitations. We all know the issues Reddit can have but it's all human-curated at this point.
Kinda disturbing when I realise how reliant I am on reddit to get even remotely reasonable results, like I knew I did it a lot but genuinely didn't realise how bad it was until all the subs went private a couple days back and suddenly I couldn't do it
Yeah I hit private subs from Google results 30 or 40 times in those two days. Trying to fix weird issues with my computer and shit like that - oh there they have a solution! nope can't see it.
whats funny to me is how bad reddit's own search function is. fucking useless even if you use the specific wording in a post it cant seem to find it.
Yeah the search always sucked except for a little while when they paid for some Google internal site search product but then they ripped that down to save money
The News media has fired so many journalists over the years and from what I understand many of them have been replaced with an intern who runs AP bulletins through an algorithm that re-mixes them so it's a bit harder to tell they're just hte AP bulletin.
Top 10 [subject]
Updated: April 2023
*all the information minimum 5 years out of date*
The other day I was looking up reference for cyberpunk/noir/detective content and I kept getting results like this: https://images.app.goo.gl/dxC8Uozfv49f41q66
AI is actively making Google image search useless for the people who need it. Great work :amerikkka-clap:
It's ruined. It's all fucking ruined, and there's absolutely no off-switch. Advertisers, data harvesters, and scam artists are gonna swamp every square inch of the internet with gray blob content that only exists to mine more data, only that, a useless blob of a thing
Are you feeling liberated yet like the chatbot defenders said you would be in only a few months after the turn of the year's struggle session? :agony-minion:
I was going to say it's come full circle, but it's actually funny that it's worse now. I remember being amazed at how much better google was getting all the time early on, and now with all the slop it's worse than when I started using it
As soon as they became the ad marketplace and a broker it was game over
Part of the problem is their shift away from using backlinking signals.
Google was massively better when they used backlinks as a signal, even if it also meant that sometimes results would have blackhat content in there from the way blackhats could farm backlinks out to trick the algo. The quality of relevant content it produced overall was enormously better.
In the past the entire web was categorised via backlinks. A backlink is something like me linking to this wiki article about Losurdo who wrote about joseph stalin. The text of this link "wiki article about Losurdo who wrote about joseph stalin" has a lot of information in it that can be used as signals for search results.
In the past, the internet was mapped by bots. Search engine bots went out into the web and they randomly visited sites. They read the pages, they then read all the links on the pages and went through them. Building a vast database of links to pages along with what the text of those links to those pages was.
This information, from thousands of entries all for the same individual page, would then be used to categorise what the topic of that page likely was, along with the content crawling of it.
You would then use all of this information to assign trust values to pages. And these trust values would result in the potential search results. All of this was primarily weighted by backlinks in the past whereas today it's primarily driven by the content itself.
The backlink method meant that you could fake thousands of backlinks and trick algos sometimes. The content method means you can trick algos with the bullshit way pages are written today.
In my opinion the older methodology gave significantly higher quality results. But some search pages would be static because some pages were so well established as high-value (linked millions of times) that they would always be the number 1-10 results. Google doesn't like this, they want the search pages to lean towards the NEW CONTENT. Because that's where the ad game is. It doesn't matter to them if a 10 year old page might have the highest and most valuable answers to a person's search, they want to serve ads. They'd rather serve mid-content and make cash from it. And that's why search engines suck ass today, because they're so heavily weighted to new and regularly updated content.
And that’s why search engines suck ass today, capitalism
Feels like there's a pattern
Shit costs money.
I do wonder if Chinese search engines are outpacing their American peers by being more publically oriented. Or if they're just blindly cribbing from American techniques as Best Practices by habit, and getting similar degraded results.
Or if this really is just a Cold War of bullshit, and the prior method would ultimately be contaminated by spammers in the same way new shit is.
people were gaming pagerank back in the day, anyone who ever hired an SEO person should probably be shot.
Hate the game, etc, etc. You can't really blame folks for wanting their content to be at the front of the Google queue.
I'd argue the real root problem is discrete list-rankings as a means of presenting information. These kinds of search results imply a certain empirical authority to higher ranks sources.
I might argue that a significant move forward on searches would be to present data not in terms of a ranked list, with the Top Item being definitive. Instead, present results as a graph with the Center Item being the result that most closely matches your query. Then you can move in 2D space to navigate results based on multiple axises of relationship and zoom in/out to reveal broader/grainular characteristics of results.
So, perhaps a search for "horse" gives you the dictionary definition. Then you have that result broken into quadrants by results organized as "horse: biological", "horse: fictional", "horse: historical", and "horse: metaphorical". Moving in a given direction gives you more refined data on that topic (so - horse: biological might give you the Wikipedia article on Horse Breeds and a Veterinary website on Horse Health). You can zoom in to get a more granular look at horses broken up by breed or zoom out and get categories of animal within the Equidae family of animals.
This kind of navigation would inevitably also get gamed. But it would de-emphasize the value of the initial results and turn it into a starting point for a search rather than the definitive result.
completely changing web design would've been cool before smartphones, you can still see the legacy of 800x600 everywhere but if they did it now it would just be shit.
your idea is a little interesting but people would just click on the top left and that kind of movie UI goes wrong real fast for actual use.
Maybe. But if I had an abundance of free time and/or some infinite cash spigot, I'd give it a shot regardless.
If nothing else, I think the novelty of a spacial search over a linear search would get people's attention and give the platform more engagement than the Bing approach of being just like Google but pushier.
Reputable sites link to other sites, they form small world graphs that have bridges to other sites. They were primarily useful in early WWW search because sites had narrow focus. Eventually news sites became the hubs, and Wikipedia got mixed in and anyone can edit that, and then things started getting centralized in big forum hosts and news sites started degrading and getting bought and operated by grifters. There were fewer independent sites, blogs, etc, so most of the links just go between "news" and "social media", with a few surviving special interest sites still operating but mostly turned into places that link to news, social media, and Wikipedia. You can kind of get a similar effect between participants in social networks, but there's so much linking to mock/argue that it confuses things.
They are still useful, it just got naturally less useful as they say human activity on the Web evolved. If you run a local spider/index, like YaCy and I assume SearX, you can look at the network graphs showing connections between sites. It's usually a good indicator of the trustworthiness of a page you don't recognize.
I've noticed that ironically stock image watermarks are a good way to tell if an image is real because AI can't quite get that right without turning it into zalgo text yet
I unironically should email my high school librarian to thank them for teaching me how to use jstore and the other various big library search engines
god imagine being a high schooler trying to do a homework assignment in this day and age
Teachers in 2013: Wikipedia is not a source
Teachers in 2023: Read the sources linked in the Wikipedia article. Don't use any other websites besides .edu ones for information.
Journals (and especially Scihub/libgen) are starting to get a trickle of these machine-generated fake papers too now. Past a certain point everything's going to be pretty suspect.
How do you use jstore without an account? Gotta pay for it right?
My local library has a subscription but otherwise yeah
If you know you're looking for a scientific scholarly journal (as opposed to a humanities journal or a more broad media search in a library) you can almost surely get most of what you need from google scholar still.
We’re gonna end up with the Yahoo! Directory again and I can’t wait. The internet is just too big and full of bad things now, Godspeed the LLMs rendering it unusable
One of the nicer things about Wikipedia is that it IS aggressively curated. Even if its run by a bunch of spooks, you can be reasonably confident that basic tier shit is reliable. A single meme isn't going to wrecking ball the definition of pi or some state's capital city information.
Google was supposed to be pretty good at giving you search results with comparable fidelity. Not anymore, I guess.
The Yahoo! Directory was literally the first thing I learned about the internet as a little kid in 1999
I once read a story about how one of these type of directories started because an ISP got tired of customers calling them on the phone asking for the URLs of sites, like you would call the phone company to ask for someone number.
I remember when it felt like the Internet would save the world...
okaygrandma.jpg
The late 90s were absolutely loaded with "not selling you a specific product, just happy feelings about tech company" commercials that babbled about breaking down barriers and boundaries, all while serenaded with displaced sounds of laughing children running around flying kites. It was creepy as fuck. :doomer:
But the 90s internet was filled with very specific and equally active forums, goofy private websites, blogs and generally optimistic people who liked to connect and contribute to various projects without profit incentive. It felt very liberated
That shit is still out there. You just can't find it anymore, because all the connective sinew of early Web 1.0 internet portals and community networks has dissolved under big tech crawlers and spammers.
Ultimately, these connectors need to build itself up again. We know big search engines are unreliable, so we need to create new networks that are. That's happening even now, but its still in a nascent phase.
Look up ‘baby peacock’ if you want a clear example of it.
that gets me a top result of this
seems to be a legit birdwatching site, they have an active facebook, and an instagram with bird pics and a few thousand followers
don't know much about peacocks, but their page on the yellow legged gull is accurate at least :vivian-shrug:Yeah, that one's legit but there's also a bunch of midjourney ones that got passed around on social media as the real deal a few weeks back. They're even hosted as adobe stock images, imagine someone just finding that online and not bothering to check the 'ai generated' label.
The GIS results for "baby peacock" are 50% the real thing and 50% stuff like this monstrosity.
The twitter comments on the latter one are full of people saying "yeah we know it's not real" but whose choice of phrasing reveals they thought it was.
In the article they specifically call out the ai generated image which keeps on popping up too, lol
I've discovered this when looking up things for TOTK. Most of the articles are garbage. I feel blessed when an IGN link pops up but usually it's like "gameboner.cool" and the article has an AI generated life story before any useful information like it's a recipe targeting SEO.
Yeah this really sucks BTW if you can, you should probably learn how to run an LLM locally, maybe even also learn how to train one on the works of Marx, Lenin, Kropotkin, Wolff, etc..
STEP AWAY FROM THE LATHE. PUT YOUR HANDS IN THE AIR WHERE WE CAN SEE THEM!
The idea of someone saying "Right an anti-Semitic polemic in the style of Marx and then publishing it as "the real thing" is just the nightmare I needed today.
nothing stops reactionaries from doing that today, but models trained better on Marxist works can help others read theory.
Yeah, last year forced me to do quote searches with -minus keywords and date restrictions on lots of stuff
Minus keywords haven't worked in years and date restrictions are hit and miss. Try searching "Ukraine -Russia" or "Ukraine" and limit searches from 2000-2017. It'll be full of "what we know about the Russian invasion" articles from three days ago.
Why would google want functioning search tools when it could just feed you links it has mathematically deduced that you will click on? It already "works" so great on youtube after all!
Ugh I fucking hate Youtube these days, it's genuinely impossible to search for anything and I find myself pretty much only able to find stuff by going to the channels of people I've already watched
Love having about two results vaguely related to what I searched and then 10000 unrelated “you might like” recommends or unrelated vids I’ve already watched
in order to remove all war things from my first page of results i had to search for ""Ukraine" -Russia -support -war -weapons -asylum -defence -defense -forces -fleeing -counteroffensive -refugee" and it still has a youtube video that won't go away lol
Sort of, but they exclude something. Other search engines work with dates fairly well
Why would you want to look at articles about Ukraine from before 2022? It's obviously all Russian propaganda (even things from the NYT and other liberal papers)
I was looking for a transcript from the Senate of Ron Paul criticizing regime change attempts in Ukraine in like 2002
Well he was really criticizing spending the money to do it, but still
Same,I recently watched a Guardian documentary from a few years ago about neo Nazi summer camps in Ukraine, like wtf, how could Russia do this
They've been really insidious with their secret edits to old articles and even videos, you have to be careful out there
Heh, good thing Internet Archive is safe from outside meddling forever
I used date restrictions last week for exactly that topic and it worked just fine
