New research from the firm Graphite found that around half of all articles on the internet are AI generated.
  • U7826391786239@lemmy.zip
    82·
    3 days ago

    seems like more than half the articles i click on anymore are shit-tier model garbage that was generated the instant i clicked the link

    no proofreading, no editing, just instant word salad for the purpose of getting ad views

    • WALLACE@feddit.uk
      23·
      3 days ago

      I was researching new vacuum cleaners recently, and found one of those websites that allegedly test and compare different ones.

      One of the entries had wording to the effect of “This vacuum cleaner may have a weight of <X>, and might have feature <Y>” etc.

      It was fucking AI generated garbage that didn’t even know anything about the vacuum cleaner so it started speculating.

      • U7826391786239@lemmy.zip
        9·
        3 days ago

        yea. for larger purchases it’s worth a trip to the library and ask them to see the consumer reports magazines for vacuums (or whatever you’re looking for). some libraries have them for in-house use only because they’re stolen so frequently

        edit: just saw you’re in UK; no idea if the thievery applies to your libraries or not

        • smh@slrpnk.net
          3·
          3 days ago

          Some libraries also have access to Consumer Reports online, accessible via the library’s network (or maybe also via a library card + link from the library site? I dunno, it’s been a while since I worked in a public library.)

        • WALLACE@feddit.uk
          2·
          3 days ago

          Never even thought of that! We have a well known household review magazine here called “Which?” but I always used to turn my nose up at the idea of paying for reviews when I can just go on the internet and find semi-trustworthy reviews for free. Looks like those days are gone now.

    • panda_abyss@lemmy.ca
      18·
      3 days ago

      If I read one more article and get to the first subheading titled “The Investigation: Hyperbolic Subheading” I’m going to delete the whole thing.

      It’s so fucking annoying to read, and you know 100% is an AI title because nobody writes that way.

    • The Velour Fog @lemmy.world
      12·
      3 days ago

      That’s pretty much what it is. Looking up song lyric meanings is a crapshoot, too.

      I was looking up “People Watching” by Sam Fender and all the AI generated articles were saying he was singing about a lover that he lost. Looked up the song on Wikipedia and it said that he was actually singing about his mentor being in palliative care as well as a criticism of the healthcare industry.

    • stabby_cicada@slrpnk.netOP
      12·
      3 days ago

      One: the original Axios article requires you to log in.

      Two: The researchers were aware of the limits of AI detectors, and tested the one they used. From the article:

      We should also take the judgments of AI detectors with a grain of salt, since their reliability is up for question. In its own testing of Surfer’s accuracy, Graphite had the detector analyze a sample of AI-generated articles and another sample of human articles, finding that it labeled human-written articles as AI-made 4.2 percent of the time — a common problem with these tools — but only mistook AI-written articles as human 0.6 percent of the time.

      And really, judging from the quality of search results these days, I would have expected a lot more than 50% of new online articles to be AI generated, so from that standpoint the article might be good news 😆

    • ekZepp@lemmy.worldEnglish
      8·
      3 days ago

      Sure. Also neighbours rumours are a very effective way to stay connected with what happen in the world and have very real news.

  • jaybone@lemmy.zipEnglish
    2·
    3 days ago

    I have created several dumb little web pages on various free hosts. They are not AI contents. They either don’t get indexed by search engines, or don’t get returned in the results. But search results return plenty of AI garbage.