Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
AI Google

AI Images in Google Search Results Have Opened a Portal To Hell (404media.co) 70

An anonymous reader shares a report: Google image search is serving users AI-generated images of celebrities in swimsuits and not indicating that the images are AI-generated. In a few instances, even when the search terms do not explicitly ask for it, Google image search is serving AI-generated images of celebrities in swimsuits, but the celebrities are made to look like underage children. If users click on these images, they are taken to AI image generation sites, and in a couple of cases the recommendation engines on these sites leads users to AI-generated nonconsensual nude images and AI-generated nude images of celebrities made to look like children.

The news is yet another example of how the tools people have used to navigate the internet for decades are overwhelmed by the flood of AI-generated content even when they are not asking for it and which almost exclusively use people's work or likeness without consent. At times, the deluge of AI content makes it difficult for users to differentiate between what is real and what is AI-generated.

This discussion has been archived. No new comments can be posted.

AI Images in Google Search Results Have Opened a Portal To Hell

Comments Filter:
  • That's the point (Score:3, Interesting)

    by Wonko the Sane ( 25252 ) * on Tuesday June 18, 2024 @11:48AM (#64558261) Journal

    At times, the deluge of AI content makes it difficult for users to differentiate between what is real and what is AI-generated.

    When true information gets in the way of a regime's goals, then any regime will attempt to disrupt the propagation of true information.

    Poisoning search results with fake content is such a tactic.

    • by ClickOnThis ( 137803 ) on Tuesday June 18, 2024 @12:06PM (#64558345) Journal

      At times, the deluge of AI content makes it difficult for users to differentiate between what is real and what is AI-generated.

      When true information gets in the way of a regime's goals, then any regime will attempt to disrupt the propagation of true information.

      Poisoning search results with fake content is such a tactic.

      I'll grant you that "any regime" will exploit AI for its benefit. But that doesn't mean that "any AI" is from a regime.

      Like any other technology, AI can be used for good or bad goals. Distortion of the truth is the problem, not the existence of AI.

    • by Rei ( 128717 ) on Tuesday June 18, 2024 @12:22PM (#64558427) Homepage

      Why exactly are you searching for "[CelebrityName] swimsuit" anyway, and why exactly do you think "the regime" is getting involved?

      • Re: (Score:1, Informative)

        by Anonymous Coward

        Why exactly are you searching for "[CelebrityName] swimsuit" anyway,

        At least read the fookin summary:

        even when the search terms do not explicitly ask for it, Google image search is serving AI-generated images of celebrities in swimsuits

        • I'll speculate that this is caused by the image-providers overloading their pages with popular (but irrelevant) google search-terms. In other words, I wouldn't immediately suspect Google of causing the problem.

          • Gaming the system, in other words. That's a major failing of modern advertising systems. Ie, Google, not a tech company but actually an advertising company.

            Remember the good old days when advertisements were curated? Every advertisement, everywhere, had to have someone approve the advertisement, whether it be on a billboard, radio, television, etc. Even an ad on a company's front door for the company's own product required somebody had to approve that. These days, most ads are automated. Google does no

            • by HiThere ( 15173 )

              It may get a massive hit to profits if it DOESN'T start curating this stuff. If there's a defamation suit, Google will be the party with the deepest pockets.

            • I think the topic was search results (not ads) but I get you. What you're describing is a recommendation system that customizes what a user sees on a page, based on demographic information about them. Companies that pay for ad placement understandably want to pay for eyeballs that might be interested.

              The technology isn't just used for ads, but also for custom pages on vendor sites, such as Amazon product pages that show you what other people also looked at or purchased. And so on.

        • by Entrope ( 68843 ) on Tuesday June 18, 2024 @12:51PM (#64558511) Homepage

          Yes, that's weasel wording by a click/rage source. If you look at the image in their teaser, the search string is garbled but ended in "swimsuit". I conclude that the narrow meaning is that they did "not explicitly ask for" modified or AI-generated images -- but that clearly did ask for swimsuit images.

          • by Rei ( 128717 )

            Exactly. They give no examples of "celebrities in swimsuits" being returned when the person isn't looking for said celebrity in a swimsuit. Their complaint is that they're not real images of celebrities in swimsuits. Which is just inane.

      • by narcc ( 412956 )

        The problem isn't that your search returned relevant results that happen to be AI generated. The problem is your kid's search for "peppa pig coloring pages" returns AI generated images best described as "naked statues of Natalie Portman, covered in hot grits".

        I can't wait for this stupid fad to be over...

    • by Alinabi ( 464689 )
      "Do not attribute to malice that which is easily explained by incompetence."
      • Malice is often the result of incompetence. If one is competent that usually don't have to resort to evil. If Hitler had actually been a decent painter...

        • For what it's worth, he was a decent painter. Plenty images of his paintings exist - They're not masterpieces or anything, but they're better than what most people could do if you handed them a paintbrush and a palette.

          Thing is, you don't get into the Vienna Arts Academy by being decent at your art.
  • Well, seeing as how I will maybe be spending eternity there, can I get a quick tour so I will know what sins I need to commit to get to the places in Hell I will hate less?
    • If you're going to hell because of one sin, why not make it 1,000 sins and be a legend.

      • oh, i am pretty sure it is not just one. Whether or not it is 1000? I think it is more like 30,000 at this point, maybe 40,000. I know i have broken at least one commandment over 10,000 times.
  • Paywalled content (Score:4, Insightful)

    by Sloth77 ( 6645172 ) on Tuesday June 18, 2024 @11:54AM (#64558299)
    Please don't link to articles that are paywalled
  • by iAmWaySmarterThanYou ( 10095012 ) on Tuesday June 18, 2024 @11:58AM (#64558319)

    Since when do real human beings repeat themselves like that?

  • Well, that's what AI is good at. [creator.nightcafe.studio]

    • by Rei ( 128717 )

      This has never happened on the internet before.

  • by Retired Chemist ( 5039029 ) on Tuesday June 18, 2024 @12:14PM (#64558377)
    So, the search programs should develop an AI to identify AI created items and block them. The AI programs can duel to death, while the rest of us get on with life.
    • So, the search programs should develop an AI to identify AI created items and block them. The AI programs can duel to death, while the rest of us get on with life.

      Do you want SkyNet? Because that's how you get SkyNet. ;-P

      • by narcc ( 412956 )

        What he's describing is a GAN. The basic idea is to have two models, one to generate things, called the generator (G), and another to discriminate between real and generated things, called the discriminator (D). The concept is very simple. We use D to improve the quality of G, and G to improve the quality of D. As G gets better at making fakes, so D gets better at identifying them.

        This works remarkably well, as you might imagine, but they will not continue to improve forever. Too much training and your

        • Yes, I get all that. My post was a joke. Maybe a bad one, and maybe not obvious.

          My point was that when AIs start to fight each other, one will "win" and then ... look for other combatants?

          • by narcc ( 412956 )

            I had wondered, but there are enough singularity nuts here that it seemed worth the (admittedly low) effort.

            I look forward to your Kumite of AIs.

        • by AmiMoJo ( 196126 )

          They could just detect images of 12 year olds with D cup boobs and 90% of the AI crap would go away.

      • No this is how you get Metal Gear REX. SkyNet wasn't developed to fight other AI.
  • The Agenda (Score:2, Troll)

    by The Cat ( 19816 )

    The nightmare scenario is if people run off and start communicating with each other where big tech can't monetize and where big government can't fear-monger.

    Up to now the business model has been to set up a tool booth and a facial recognition camera on every road where communication between individuals might take place.

    The future business model will be to poison all information, visual or otherwise, with the suspicion that it was generated by AI. That will artificially create a need to "protect" people from

    • pardon, but I am curious, to which, "brutal consequences for non-compliance" are you refering?
    • This... isn't helpful. There's plenty of tyranny, oppression, corruption, and complicity that's verifiably happening in the real world without having to drum up conspiracy theories. And make no mistake: AI generated content is a headache for the powers that be just as much as it is for us. Sure, they'll exploit it for their own influence campaigns, but that's a situation forced upon them, not some Machiavellian plot they devised.
  • Is anyone surprised?
  • by DigitalSorceress ( 156609 ) on Tuesday June 18, 2024 @12:53PM (#64558517)

    AI image generation is pumping out so much garbage that it's going to start feeding on itself and I'm awaiting the inevitable Hapsburg (heck we kind of started there didn't we?) then Cronenburg and finally if allowed to keep consuming the output of AI we'll get to whatever image generation version of Gray Goo is..

    Sadly, I'm sure this won't stop the loss of jobs for graphic artists and other artist types...

    I used to enjoy dystopian cyber future genres back in the 1980s - now that I'm living in one - not so much.

  • Where's the Doom Marine when you need him?

  • by nightflameauto ( 6607976 ) on Tuesday June 18, 2024 @01:00PM (#64558545)

    Are nude children hell? Are nude celebrities hell? Are nude celebrity children hell? Are nude children with celebrity heads on them hell? I'm just trying to set my gages for what, exactly, is hellish about seeing nude children. Anybody that's ever given a kid a bath or changed a diaper is familiar with hell now? Great.

    In the coming crapflood apocalypse, this is pretty small potatoes. It's people looking for smut and getting a different level of smut than they wanted. Boo hoo. Ain't like that part of the crapflood is gonna cause massive damage to anything other than some fragile psyches that were clearly already verging on broken.

    I'm more concerned about how quickly the crapflood is rising. Now that we have AI everything spouting nonsense everywhere, it's only a matter of time before we get crapflooded to the point that nothing on the Internet will have any reality behind it. I mean, we were already doing a pretty good job of that ourselves. Now we have AI to help us crapflood more efficiently! Our information age is starting to grind to a halt before our eyes. Gonna be one heck of an interesting ride over the next few years, while the completely out of control corporations do whatever they want to the rest of us, as our government sits back and goes, "*BLINK*BLINK* Um, what?"

    The Internet had its time. Now, as with all things, that time has come to a slow, grinding, churning, toxic end. RIP information age. It was fun while it lasted. What a time to live through.

    I welcome our disinformation age and its AI driven fountain of shit. Bow to the new masters, controlled by the old masters, spewing nonsense over all of humanity.

    • ... nude children ...

      This has been considered 'bad' behaviour for 30 years: Which is why we have mothers, mostly in Florida, complaining about children seeing Drawings of nude children, and also, statues of nude men, and also, the ugly truth of what she did to turn into a mother.

      It's "think of the children" censorship claiming children are saved by ignoring normal, daily behaviour.

      You probably can't remember the last time you saw a nude child on Tv.: It's reached the point of subtle sexism, where little girls aren't on Tv

    • Gages are hell. Gauges are fine though.

    • by AmiMoJo ( 196126 )

      In some parts of the world, images of AI generated children in the nude can get you arrested. From the point of view of the search engines, they probably don't want innocent searches for some pop star to return AI generated porn of them with a child's face, for the sake of the user as much as their own reputation.

      I'm not a child psychologist, but people who are child psychologists say that even AI generated porn of real children can be psychologically harmful to them. Apparently "get over it, it's not real"

  • im disappointed, wheres my fire and brimstone???
  • Slashvertisement FTL. Go cry in your church that you likely need to reread the book from yourself.

  • So you're telling me that the US Presidents didn't actually have mullets [nypost.com]?

  • I'll admit, I don't always appreciate the sight of nude people...it rather depends on their physique...but "Hell"???

  • THE point of Google is fast, accurate searches. That's it. That's how their brand name became a common verb (as in "google it"). Anybody could go to Google and get a fast-loading uncluttered page with basically the corporate logo and a search box to type into. While the search results were never perfect (how COULD they be when pages are changing every moment all over the planet?) they were as good as a person could get for free.

    Now, because some moron somewhere at Google has the AI bug and (like the little

    • by Rei ( 128717 ) on Tuesday June 18, 2024 @08:02PM (#64559575) Homepage

      Virtually none of the errors that people were posting all over were "computer invented". They were human invented. Eating rocks? Humans. Glue on pizza? Humans. Muslim presidents? Humans. It was a RAG summarization model, tasked with summarizing the top search results. And by and large it did its job properly.

      The problem wasn't with the AI model; the problem was with the very concept. The notion of summarizing the top search results without any attempt to assess their credibility, sarcasm, trolling, etc, and then presenting it up front as an authoritative answer, is idiotic. You can train a model to assess those things, but they did not, whether due to wanting to keep the model lightweight (so it's fast and cheap to run), or just a sheer lack of insight on their part.

      Amusingly, at one point I presented Google's plan to ChatGPT and asked ChatGPT what it thought of the plan. It noted a few positives but then launched into a long laundry list of potential negatives (not just how bad or dangerous the results could be, but the brand/reputation damage that would result from it). Every single one of the potential negatives it listed played out. Aka: if Google execs had bothered to ask a good LLM, it could have told them how dumb their idea was. ;)

  • It's going to get much worse. Social media already has negative value, and search is not far behind. Both will get worse until big tech embraces cryptographically signing all content. Finally, an actual use case for a distributed crypto ledger. Every photo you take, and every document you write, needs to be signed. Anything not signed is not trustworthy.

    Adding metadata that says "This is AI generated" does not count. It will just be removed, and there will be tools that simply don't do it.
    An image signed b

  • ... celebrities made to look like children.

    Disney's Tinkerbell, is a baby-fied face on a beautiful woman's body. Japanese anime, with it's big eyes, is also reminiscent of a baby-fied face.

    • those are drawings; no humans were abused or violated in their drawing

      how can you draw any connection to actual people being debased?

  • Now that would not only need an AI warning.

  • Lately the whole ranking is littered with these fake webshop pages.

Someday somebody has got to decide whether the typewriter is the machine, or the person who operates it.

Working...