Comment on Firefox is Getting a New AI Browsing Mode

Lumidaub@feddit.org ⁨3⁩ ⁨weeks⁩ ago

• over_clox@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  Prolly like if you look up something that doesn’t actually exist, it’ll find you results anyways.

  This ought to be fun for Rule 34…

  source
• Meron35@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  There is an AI browser war going on right now, see Comet (Perplexity), Atlas (ChatGPT), Claude AI Agent from Chrome etc.

  They work by letting the AI continuously see everything your browser can see, such as your emails, banking details, financial habits, online shopping accounts, etc.

  By doing so, they promise to be better digital assistants, so that you ask just ask the browser to do tasks such as online shopping, booking holidays, etc.

  Even ignoring the severe privacy concerns, AI browsers are significantly prone to prompt injection. That is, any random webpage with hidden text can override the instructions you give it to carry out malicious attacks.

  source

  • Griffus@lemmy.zip ⁨2⁩ ⁨weeks⁩ ago

    Thanks for the explanation. I hate it.

    source
• riskable@programming.dev ⁨3⁩ ⁨weeks⁩ ago

  Today, when you search for something you get a handful of ads, SEO-sponsored bullshit, and maybe the 6th link will be what you were actually looking for.

  When you tell the AI agent (in your browser) to search for something, not only do you get the most relevant results (because you can make your prompt vastly more specific and detailed), you completely skip all that other stuff that you didn’t want.

  I’ve been saying for some time now that AI is going to kill free search engines because it’s such a better way to search for stuff. Free search engines like Google and SEO-optimizing companies are hindrances to efficient browsing and drowning the web in bullshit. Poisoning search results.

  An AI agent will skip past all that stuff and give you just what you want; you never see any ads!

  source

  • seathru@lemmy.sdf.org ⁨3⁩ ⁨weeks⁩ ago

    An AI agent will skip past all that stuff and give you just what you want; you never see any ads!

    You’re wondering, “Why is my cat sneezing?”

    While you ponder that, why not treat yourself to something delicious? At Carl’s Jr., our Charbroiled Burgers are made with 100% Angus beef, grilled to perfection, and packed with bold flavors. Treat yourself today—you deserve a meal that satisfies!

    Now, back to your question: Sneezing in cats can be caused by allergies, respiratory infections, or irritants. It’s always a good idea to consult with your veterinarian to ensure your feline friend is healthy.

    Let me know if you have more questions!

    source

    • riskable@programming.dev ⁨2⁩ ⁨weeks⁩ ago

      You’ve obviously never used an open source AI model (running locally on your PC) if you think that’s how it’d go.

      source

      • seathru@lemmy.sdf.org ⁨2⁩ ⁨weeks⁩ ago

        Likewise, a properly set up and locally hosted searx instance should not be returning ads and “SEO-optimized bullshit”. You’re comparing “free” corporate owned services to privately owned ones. OFC corporate owned is going to feed you ads.

        source

        • -> View More Comments
  • Lemminary@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

    That’s assuming the AI won’t look at the results and still make shit up. I’ve used AI-assisted search and i know that it’s not reliable.

    source

    • riskable@programming.dev ⁨2⁩ ⁨weeks⁩ ago

      Ok how would that work:

      find me some good recipes for hibachi style ginger butter

      AI model returns 10 links, 4 of which don’t actually exist (because it hallucinated them)? No. If they didn’t exist, it wouldn’t have returned them because it wouldn’t have been able to load those URLs.

      It’s possible that it could get it wrong because of some new kind of LLM scamming method but that’s not “making shit up” it’s malicious URLs.

      source

      • Lemminary@lemmy.world ⁨2⁩ ⁨weeks⁩ ago

        If they didn’t exist, it wouldn’t have returned them

        And yet I’ve had Bing’s Copilot/ChatGPT (with plugins like Consensus), Gemini, and Perplexity do exactly that, but worse. Sometimes they’ll cite sources that don’t mention anything related to the answer they’ve provided because the information they’re giving is based on some other training data they can’t source. They were asked to provide a source, but won’t necessarily give you the source. Hell, sometimes they’ll answer an adjacent just to spit out an answer–any answer–to fulfill the request.

        LLMs are simply not the appropriate tool for the job. This is most obvious when you need the specificity and accuracy.

        source

        • -> View More Comments
  • JandroDelSol@lemmy.world ⁨2⁩ ⁨weeks⁩ ago

    ah yes, because there’s no way the programmers won’t induce bias in their AIs

    source

    • riskable@programming.dev ⁨2⁩ ⁨weeks⁩ ago

      If you don’t like your current open source AI, just use a different one or an embedding that works around whatever bias you don’t like. Maybe open a ticket?

      source