A place to discuss the news and latest developments on the open-source browser Firefox
Just today a number of my searches had me wondering whether AI slop generators were generating articles and suitably titled websites in response to the searches I was making on my search engine.
I don't know how they'd do that without being able to snatch search terms I make to my search engine (DDG, with its AI features disabled), just don't find it plausible that anyone is bothering to generate hundreds of sites on extremely narrow aspects of obscure topics, even if semi-automated.
No, but I've had A LOT more cloudflare 'prove that you are human' boxes. And I had an infinite loop captcha. I don't have much in the way of extensions, and I wasn't using a vpn
It's not just tech. Gardening, DIY, cooking, and similar popular subjects have been completely destroyed by this crap. If I see an AI generated header image or thumbnail I immediately backpedal now because I assume that means the text is bullshit too.
The example stuck in my memory now is when I was trying to read about watermelon growing times and the article said they flower a week after germination.There's now frequently this, "oh GOD DAMN IT *close tab*" moment when you realize it's actually total slop. Like, "oh so this article is BULLSHIT bullshit."
I have been worried about this for months. The moment of clarity for me came when I was looking up how to tan a hide fur off. I have done fur on and seen a dozen videos on it so I knew what I was looking at. Found this great looking site. Step by step. As I'm reading, something seems off. Maybe translation issues? The order here seems... Repeatative?
Then I look at the pictures.
First picture showed a hide being made into a drum. Ok, not exactly what I'm doing but yeah maybe similar process.... Next picture is of a dudes abs, with a tan line.... (Tanning the hide)
Next picture is a head table for a wedding... (Decorating)
It was surreal. And it is surreal. We are now returning to a time when we can't access information easily. Not because it isn't there, but because it's crowded by misinformation and half information.
If I'm looking for information that doesn't need to be super recent/up to date, I just limit my search results to before November 2022 at least.
It's the "nuclear option" but it's worked for me a handful of times.
Now just wish DDG let me have that filter for the images tab to make finding new wallpapers easier for me...
The problem predates the AI craze. SEO slop started snowballing before.
Highly recommend reading this: https://www.wheresyoured.at/the-men-who-killed-google/
Non-Google search engines are kinda the in-progress solutions, as the sheer volume of stuff to block is likely intractable. Google has the power to fight it themselves, but, well, see the writeup...
Reject hypertext, return to gophe
Gemini is nice. It's got some serious QOL improvements over Gopher too, for those of us that like a little formatting or structure.
With how much SEO spam/slop that comes up in search engines, it would probably just be easier to whitelist trusted websites
Yeah we need to go back to a curated list of content like back in the days of directories.
Hi, what is SEO
Search Engine Optimization. Basically gaming search engine indexing algorithms so that your content appears more "relevant" (read: crammed full of as many keywords as possible) and thus higher up on search results, usually at the expense of having, you know, actual content worth reading.
I noticed a while ago that the search results were different. Then I noticed that I don't seem to be able to find what I need anymore
Yeah, it's the kind of thing that in utopia would actually help search engines and users find relevant pages, but under capitalism becomes "hey, listen! look at ~~me~~ my ads!"
chances are, it will have a table of context and generally low quality writing
Don't forget repeating the question over and over again. If my question gets rephrased four times in the first four sentences, it's a good sign that I'm reading AI slop and there's no actual answer in the pages and pages of text.
Recently? It's been happening for years.
Every one of these websites looks the same too. Like a mix between a blog and a wiki, and always in FAQ form. You can tell that they're AI-generated because the questions will seem related to each other but the answers often aren't.
And they often have the same question worded slightly differently three or four times in the first paragraph.
Good idea
I use uBlacklist with this filter and that generally keeps the repeat offenders at least out of image search, but clearing out every SEO-spam print-on-demand mimc-site was already a game of whack-a-mole before consumer LLMs became a thing; I imagine now it'd be like playing whack-a-mole with a hydra. Still, it does at least help.
Search engines need to start using AI detectors and drop the ranking way down when any significant portion of the page is AI generated.
AI detectors are massively flawed. They have terrible accuracy and have high numbers of false positives. Especially over short bodies of text like parts of one page
replying so i remember to check back
this is such a great idea if it isn't a thing yet, there are so many of these slop pits
Tried looking up how to bake wholemeal bread with just plain wholemeal flour and white bread flour (I can't buy wholemeal bread flour at my supermarket). AI really is everywhere and it's making the internet useless.
Guess I could've done the before:202X trick.
Increasingly find that search engines ignore instructions to filter by date or site, which coupled with ignoring all operands will kill off their utility entirely.
I guess we're going back to having to look up stuff on physical books. I love the future!
Full circle
If you can find it, I keep a small bag of straight-up wheat gluten and I add a spoonful or two when I want to make stronger flour. A small bag lasts forever and a little goes a long way.