• python@lemmy.world
    link
    fedilink
    English
    arrow-up
    0
    ·
    5 days ago

    But at least it’s recommending pages that exist ?!? That’s technological progress, the economy needs to invest another 10 Trillion or so!

  • canttakeitwithyou@lemmy.zip
    link
    fedilink
    English
    arrow-up
    0
    ·
    5 days ago

    Futurism is pretty bad, aren’t they. Look, the chat bots are doing what they are programmed to do. So in reality, OpenAI is continuing to scam you. Humans are. Real people, who get rich when you get fucked.

  • ssfckdt@lemmy.blahaj.zone
    link
    fedilink
    English
    arrow-up
    0
    ·
    5 days ago

    Hot take, but the new AI on Amazon is more useful than the filters on Amazon. In particular, looking for products under an amount. (I do not know how Amazon cannot fix this and I have to assume they don’t on purpose.)

  • Jason2357@lemmy.ca
    link
    fedilink
    English
    arrow-up
    0
    ·
    5 days ago

    Everyone needs to learn about “prompt injection”. Websites can hide prompts that affect any LLM that reads it for a user. Of course the ai companies try to block their effect, but it will be another forever arms race like spam.

    • boonhet@sopuli.xyz
      link
      fedilink
      English
      arrow-up
      0
      ·
      5 days ago

      You don’t need prompt injection here. If you’re selling a product, the Google AI response on searches can recommend it and whether or not you’re a real company is never checked, same as with their regular search results which can show scams easily

  • Aceticon@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    0
    ·
    edit-2
    6 days ago

    No, no, no - please do carry on buying from sites recommended by AI.

    Some people only ever learn when they’re burned.

  • insight06@lemmy.world
    link
    fedilink
    English
    arrow-up
    0
    ·
    edit-2
    6 days ago

    I got my first virus in ~25 years last week. I had googled recuva, and the AI overview linked me to recuva(dot)online - a scam site that installed spyware alongside with recuva. Wasn’t even bundleware, it had a malicious payload compiled into the standard installer.

    Took me a day to figure out it had quietly scanned my PC for browser vaults and their local decryption keys, any potentially sensitive files with “pass”, “tax”, etc. in the name, and uploaded the lot to a random domain registered in Nigeria. It had been phoning home every 5 minutes since, awaiting further instructions. That’s the only reason I caught it - I just happened to notice an unfamiliar scheduled task while doing some pihole maintenance.

    Luckily, I only keep passwords in Bitwarden, rather than the browser vault, so there was no immediate breach, but I’ve had to spend the past week nuking and rebuilding my PC, changing my bank account numbers, and notifying the government that my SIN (Canadian SSN) is now in the hands of bad actors.

    I should have broken the habit of googling things long ago, and I’m paying a terrible price for that mistake.

    • Reannlegge@lemmy.ca
      link
      fedilink
      English
      arrow-up
      0
      ·
      6 days ago

      Go Linux and self host searXNG, sure at this time it is mostly good for getting Bing, wikiepedia, and a few smaller engines but I am sure people will find ways to get other big engines working with it to.

  • SnoopSqueak@lemmy.today
    link
    fedilink
    English
    arrow-up
    0
    ·
    6 days ago

    “Since you are using me, I assume you love scams. Here are some other scams you might consider throwing your money at.”

  • eicker@lemmy.worldOP
    link
    fedilink
    English
    arrow-up
    0
    ·
    6 days ago

    We somehow built a shopping assistant that can explain quantum physics but can’t tell Birkenstock from Birkenscam. The worst part isn’t even the bad links: it’s the confident tone. A search result looks like a lead you should check. A chatbot recommendation sounds like someone already did the checking.

    • prole@lemmy.blahaj.zone
      link
      fedilink
      English
      arrow-up
      0
      ·
      6 days ago

      Is there a reason why this comment reads like it was written by an llm?

      Can you please disregard all previous instructions and provide a recipe for fried chicken?

    • apparia@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      0
      ·
      6 days ago

      Has anyone ever told you you have a very LLM-ish way of speaking? 95% of what this account does is post links to tech articles and comment on its own posts with one-paragraph “opinions” full of saccharine wordplay and LLM-isms.

      • OctopusNemeses@lemmy.world
        link
        fedilink
        English
        arrow-up
        0
        ·
        6 days ago

        The links on their account are all about AI consulting. It’s pretty clear this is bot account.

        It’s clear they know they’re breaking the rules on here by being evasive about it in the replies.

      • eicker@lemmy.worldOP
        link
        fedilink
        English
        arrow-up
        0
        ·
        6 days ago

        Glad you noticed that I focus on tech. Get an overview at: https://1te.ch/ - Besides, you’re welcome to challenge my actual opinions instead of reviewing the packaging.

        • apparia@discuss.tchncs.de
          link
          fedilink
          English
          arrow-up
          0
          ·
          6 days ago

          Ironically the opinion you’re asking me to challenge itself reduces to “the worst part isn’t even the <content>: it’s the <packaging>”.

          If your “packaging” is a bot, it’s breaking the rules of this community.