AI images: hide. Yeah, sure.

I guess old printed encyclopedias are much MUCH better than this shit for your child's homework. Ugh, this is so disturbing.

(ノಠ益ಠ)ノ I∀

all 68 comments

sorted by: hot top controversial new old
[–] 99 points 4 months ago* (last edited 4 months ago) (9 children)

Just go to a source and search there directly instead of using a search engine. AI has made search engines annoying as fuck lately. Wikipedia

  • source
  • hideshow 9 child comments
  • [–] [S] 29 points 4 months ago (7 children)

    Yeah, that should be how it's done. Must teach that to children though, this is so brain-hurting.

  • source
  • parent
  • hideshow 7 child comments
  • [–] 0 points 4 months ago (5 children)

    Teach children: the most trustworthy source on the internet is Wikipedia /s

  • source
  • parent
  • hideshow 5 child comments
  • [–] 12 points 4 months ago (3 children)

    It is genuinely one of the most trustworthy sources.

    Back in the day anyone could make edits. But these days they first have to be reviewed by peers.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 4 months ago (2 children)

    Yeah, but if you think Reddit moderators are bad, wait til you find out about the opinionated gatekeeping mess of censorship and harmful idealism that is the Wikipedia moderation.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 4 months ago* (1 child)

    What trustworthy source doesn't have opinionated gatekeeping mass censorship?

    That's pretty much required to filter out the bullshit. You can't avoid there being a few false positives

  • source
  • parent
  • hideshow 1 child comment
  • [–] 10 points 4 months ago

    I feel like AI is creating a renewed old way of browsing. Find info about some project in a forum(lemmy) go to the actual website, not social media, not google just plain old type the URL and go directly to a human made site.

    Distrust in corpo and AI might just bring back websites made by humans. Personal blogs, forums, product pages etc...

  • source
  • parent
  • [–] 40 points 4 months ago (4 children)

    Yeah sadly the No AI filter isn't 100% accurate, it does try though and is better than no filter IMO

  • source
  • hideshow 4 child comments
  • [–] [S] 22 points 4 months ago (3 children)

    It probably is better, but the sources get worse everyday. Terrible.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 20 points 4 months ago* (5 children)

    Don't know how they do it, but it is definitely worth your time to mass report these images. Maybe take about 10 minutes and just repirt at many as you can. That's what I do. Probably doesn't do much, but if enough images are flagged, I'd hope they take a look at the website and rule whether to block that site completely. It's what I would do if I were them.

    Edit:

    Saw in a reply that you have reported them. I personally think we should strive to create a society where genAI is consistently mass reported as what it is. And that starts with people like us.

  • source
  • hideshow 5 child comments
  • [–] [S] 5 points 4 months ago (4 children)

    To be fair, it seems a lot of these images come from the same 2-3 sources. They just pollute the internet with their slop. I believe they have many other slop images for everything so blocking them would probably cut a lot of these results. They might not want to completely remove these sources because they actually have useful images as well but at this point DDG should just cut them out because they're heavily infected.

    If this goes on, we probably end up with a curated human-centric small and limited internet at some point and it will be hard to protect it. That probably also means far from ads and SEO-based internet too, since that's what started the pollution in the first place.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 2 points 4 months ago (1 child)

    Wasn't Google actively pushing SEO, too? Like filtering information pollution wasn't a huge part of a search engine's job.

  • source
  • parent
  • hideshow 1 child comment
  • [–] [S] 2 points 4 months ago

    They were, and initially it was supposed to be a good thing (like every other thing Google takes on and going forward). They made it kinda mandatory if you want to be seen on Google. After Google started pushing ad business too, it became "ads inside SEO". It was not an optimization anymore.

  • source
  • parent
  • [–] 32 points 4 months ago (2 children)

    Apparently, the "hide AI" only blocks some domains that are known to be generators of the images. I did a "waterbear" search and noticed that a number of the slop is on stock.adobe and storage.googleapis, which immediately makes them "clear" for the search

    I still hate DDG for leaving the "show ai: yes" as the fucking default. Fuck them

  • source
  • hideshow 2 child comments
  • [–] 6 points 4 months ago* (last edited 4 months ago) (1 child)

    I tried the search on both Ecosia and DDG using the HUGE AI Blocklist (linked elsewhere on this thread) and Ecosia seems to filter out the stock.adobe and googleapis images whereas DDG still lets them through.

    What I want is a filter that can rid me of LLM-generated text on websites. I HATE looking for Linux-related stuff and getting that "high schooler giving a presentation" formatted crap that opens up telling me about why I'm seeking the information I've clicked the link for and then summarizes it all at the end.

    EDIT: As I scroll a little further on Ecosia I see more obvious "AI" generated images... oh well. At least DDG lets you flag them.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 40 points 4 months ago (4 children)

    DuckDuckGo did a site-wide poll about 6 months ago asking their users if they wanted AI or not. It was on their homepage. 90% answered "no".

    What did they do in response? ... nothing. It didn't fit their preconceived notions and apparent agenda, so they've just completely ignored it. The customer is always right, except when they don't want AI, then they're just wrong.

  • source
  • hideshow 4 child comments
  • [–] 14 points 4 months ago

    It probably only filters for the machine readable markings that will be required in the EU starting next month.

    But as soon as someone crops it in Photoshop it is indistinguishable from digital art made by humans

  • source
  • [–] 26 points 4 months ago (3 children)

    I was doing some arts and crafta with the girls, and i decided to use my photo printer to print them out some puctures of animals, wince they love animals. One of them said i want a funny bunny. So i searched for funny bunny. Holy shit, the result has absolutely disgusting to look at. Just animals were okay, but all it spat out was some boomer ass ai slop of "animals" i hate what the internet has become.

  • source
  • hideshow 3 child comments
  • [–] 13 points 4 months ago

    For legit and usable (CC licensed) images, I stick with Wikimedia Commons. Haven't seen any horrors yet.

  • source
  • [–] 13 points 4 months ago

    Recently did an image search on regular DDG, no AI filter, and was astonished by how much AI slop I got. Like, at least half the results were slop. The filter isn't great, but it's better than nothing.

  • source
  • [–] 13 points 4 months ago

    That's because there's no simple way to scan billions of images to determine their provenance.

    Even glancing at your screenshot what's to say a lot of those aren't older artistic digital images?

    This can be a fun learning example for you and the kid both. Download some massive packs of images online. Millions. Lump all of them into one single folder. Now sort them for common features, which were made with cameras, which were made digitally by a person, which are AI, etc.

  • source
  • [–] 12 points 4 months ago (6 children)
  • [–] 12 points 4 months ago

    I'm suspicious of the whole "noai" thing, especially when it is championed by companies with presence in AI. What wouldn't these companies just capture the noai search revenue to overshadow alternatives, continue to invest all their profits into AI, then kill noai and go full AI when the ideologically pure alternatives have fallen?

  • source
  • [–] 11 points 4 months ago

    Good job Adobe (upper right corner) for selling that super scientifically correct stock photo

  • source
  • [–] 7 points 4 months ago

    In this case, you’re seeing “artistic interpretations” based on tardigrades? They do this with exoplanets too.

  • source
  • [–] 7 points 4 months ago (2 children)

    Labels should actually mean something.

  • source
  • hideshow 2 child comments
  • [–] 2 points 4 months ago (1 child)

    Yeah, right out of the gate, we needed to make all diffusion do some kind of tell. Even if we started it today, it would be too far gone.

    Our only hope at this point is that good diffusion is so expensive that when the bubble bursts, no one is going to go to the trouble of making it.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 7 points 4 months ago (7 children)

    I'm not sure I understand, are you thinking they are all AI? At least some of them are not - I recognize them from the before times. Tardigrade images have always been goofy looking.

    The adobe stock images are AI though. It's probably a safe bet to avoid stock image sites although Alamy seems to have some good ones. Adding the photography filter helps a bit too, but I still got some stock images used on other websites.

  • source
  • hideshow 7 child comments
  • [–] [S] 15 points 4 months ago (6 children)

    Of course not, but a lot of them are, which is not good enough for No AI search. Yeah, stock images got worse, they shouldn't be called "stock" anymore. We almost got to the dead internet part and it came so fast, it's mind boggling.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 5 points 4 months ago* (last edited 4 months ago) (3 children)

    Fair, the adobe ones are labeled as AI if you click through to the link so that's a pretty big miss in the search.

    Edit: I retract my misinformed comment, I hadn't looked at how the filter works.

    Stock images have always included bullshit renders of microbes so maybe that's just something I take for granted.

  • source
  • parent
  • hideshow 3 child comments
  • [–] [S] 4 points 4 months ago (2 children)

    If it's this obvious then it's on the search engine.

    It's used to be fun to look at stock images. Now it's just disturbing.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 8 points 4 months ago (1 child)

    I don't agree with that take.

    https://duckduckgo.com/duckduckgo-help-pages/results/how-to-filter-out-ai-images-in-duckduckgo-search-results

    We rely on publicly available lists to filter out AI-generated content, like those provided by uBlockOrigin & uBlacklist Huge AI Blocklist, an open-source blocklist, manually curated by project contributors.

    This is an open source block list, which is manually put together by volunteers. It's not even reading metadata, presumably because that's too resource intensive (or proprietary/unreliable).

    Gemini was able to tell that was AI because it was able to read the image ID water mark and associated it with how Adobe sequences AI generated images. The image ID is in the file name, but if we wanted to catch it that way the filter would need to work on logic, not just keywords or site lists. Programming a filter to search the file names to identify if it belonged to the sequence would be more complex.

    Alternatively, if you clicked though to the image you are immediately told if it is AI. Whenever I search for microbe photos I always click the source to determine if it's a render or a stylized electron microscope capture, and I figure it's always been good practice to get the image from the source to verify it's not totally bullshit either in terms of the the search pulling up something unrelated or the image itself being misleading or even a humorous misrepresentation/meme.

    I think it's important to understand how the tools we rely on work.

  • source
  • parent
  • hideshow 1 child comment
  • [–] [S] 5 points 4 months ago

    I see. If it works on manpower that means we cannot catch up easily, which sucks. And it's possible that it will be overwhelming at some point, if not soon. I hate how internet become finding a diamond in garbage hills. And when you finally find the diamond, it can still be fake since we cannot tell at first sight. :/

    Anyway, this is a real tech fight and we are in the middle of this. It seems curated internet what we'll need from now on.

  • source
  • parent
  • [–] 4 points 4 months ago (2 children)

    Hiding AI images works pretty well on my end.

  • source
  • hideshow 2 child comments
  • [–] 2 points 4 months ago

    yea, just because you prompt "noAI" doesn't mean it won't be sneaking into your results. I've even found shitty AI generated articles like from geekstogeeks sneaking inside marginalia search. which is just a query database of websites, no recommendation algorithm.

    Search engines have gone to shit, your only path for reliable information is wikipedia.

  • source
  • [–] 2 points 4 months ago* (4 children)

    But I mean, really, to be fair, Tardigrades sort of do look like AI slop; silly nature!!!. Especially real electron microscope pics of them (https://www.livescience.com/61974-why-tardigrades-are-awesome.html)

    Also you asked for "waterbear"* It's like if you asked for duckotterbeaver and expected a non-AI platypus result ;-)

    * yes I know waterbear is the real nickname of Tardigrades and yes "no AI" should still mean no AI, and yes even a Tardigrade search results in SO MUCH AI ... but like, see my first point hahah

  • source
  • hideshow 4 child comments
  • [–] [S] 2 points 4 months ago (3 children)

    It's not on users though, it's on the search engine. Seems they can't catch up sadly.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 4 points 4 months ago (2 children)

    oh totally 100% !!!

    I just found it amusing because #waterbears

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 4 months ago