If they were just talking about Reddit, I’d assume something dodgy was going on connected with the IPO. But Quora is supposedly back from the dead too… Am I missing something glaringly obvious here?

  • Otter@lemmy.ca
    link
    fedilink
    English
    arrow-up
    70
    ·
    edit-2
    8 months ago

    Advertisers are probably paying more content farms to astroturf it though.

    Yup, in fact we just banned ~13 accounts tonight from a subreddit I’m still involved with. That’s just the ones we identified, and it’s only a medium sized subreddit

    A user noticed that the responses to a post sounded a little off and reported it. Turns out there was a network of bots using generative AI to mix real academic advice (ex. “Go talk to the advising office”) with occasional subtle advertisements (ex. “I recommend using grammarly and (advertised service)”.

    Once we caught on, we looked through the history of those accounts and gathered as many as we could identify and banned them all.

    I don’t think this is Reddit’s doing, and they’re usually good about banning spam bots site wide once a mod report is made. Still, they benefit from increased activity and they have an incentive to do less of that. It was also much harder to notice the problem because of the AI generation. If a user didn’t explicitly report it, I probably wouldn’t have noticed

    • Grimy@lemmy.world
      link
      fedilink
      arrow-up
      39
      ·
      8 months ago

      I highly suggest you ban what the were advertising and not just the account.

      If advertiser’s realize the shady bot farms they deal with are causing any comment that mentions their product to be automatically deleted, they will stop.

    • nilloc@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      29
      arrow-down
      2
      ·
      8 months ago

      This is going to be the Idiocratizing of the internet. AI is going to be training in itself with these unidentified posts and get dumber and dumber.

      Let’s hope no one lets it have access to anything important…

      It feels a little like how steel from before above ground nuclear testing, called low-background (or pre-war) steel because it isn’t contaminated is prized for building some sensors.

      Pre AI information need to be preserved, otherwise we might not really know if the info we’re seeking is fact based in any way.