DDG is now offering free/private AI chat using several models.

  • kinsnik@lemmy.world
    link
    fedilink
    English
    arrow-up
    84
    arrow-down
    2
    ·
    edit-2
    8 months ago

    If they are using GPT-3.5 and Claude, that means that they are sending the chats to Open AI and Anthropic, right? How can they assure that the chats are private and not being used in training if they don’t control what other companies do?

    Edit: ok, they claim to have agreements with them to delete chats within 30 days and they hide the user IP

    • TheFeatureCreature@lemmy.world
      link
      fedilink
      English
      arrow-up
      73
      arrow-down
      3
      ·
      edit-2
      8 months ago

      DDG’s classic “Trust me bro” privacy policy.

      I don’t dislike DDG and I do use it, but goddamn I’d love to see a public audit of their privacy claims. DDG is closed source and they’ve only ever given Their Word TM about their claims. The privacy community puts a lot of faith in DDG despite not being able to test anything it says.

    • TAG@lemmy.world
      link
      fedilink
      English
      arrow-up
      25
      arrow-down
      2
      ·
      8 months ago

      That is how DDG search works as well. They take your search query and send it to a regular, data harvesting search engine. The engine does not see your IP address and cannot track you with a cookie but they can monitor the search queries of DDG users in aggregate.

      • Squizzy@lemmy.world
        link
        fedilink
        English
        arrow-up
        2
        arrow-down
        3
        ·
        8 months ago

        Ddg doesnt work for search anymore, completely compromised giving the same results as google and others, exact same bubble you are in. I dont want personalisation

        • lud@lemm.ee
          link
          fedilink
          English
          arrow-up
          1
          ·
          8 months ago

          It works very well for me still. I haven’t noticed anything weird.

          • Squizzy@lemmy.world
            link
            fedilink
            English
            arrow-up
            2
            ·
            8 months ago

            Hooefully it stays like that for you. For me I use it on all devices but they are all independent. If I search a random name it will find doctors etc. close to me with that name, down to my town. If I turn off location I will get a few general results and then back to localised.

            It is incredibly annoying, especially when doing research as I cannot get out of the localised loop. Recently I needed a legislative policy from France but could not remember the name of the policy so I searched for the intention of the policy and got local results by local papers and local politicians discussing similar legislation. I could not get to the French results without using a VPN and I got very different results then.

            There is also the concern then that ddg are sending my IP with the query. So I get bad results and no privacy.

    • Cheradenine@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      18
      ·
      8 months ago

      They say they proxy the requests. Deleting the chats after 30 days is meaningless, the data has already been utilized.

  • Lvxferre@mander.xyz
    link
    fedilink
    English
    arrow-up
    27
    ·
    8 months ago

    The good: usability. Bot answers are easy to access but still very distinct from the actual websearch. And if you don’t like bot shit, it’s easy to ignore. (I found some use case for it, but I get that plenty people dislike this.)

    The bad: as others mentioned, even if you trust DDG, you have no reason to trust Microsoft’s vassal OpenAI and its warring feud equally infested by tech bros Anthropic.

    The ugly: the financial motivations behind the move are unclear. What is DDG trying to get from that? What is its business strategy?

  • Varyk@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    10
    arrow-down
    1
    ·
    8 months ago

    Whoa, awesome. I just tried it and copilot, which I did like, has been supplanted.

    I’m getting way more accurate answers.

    • Rolando@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      ·
      8 months ago

      I tried out ddg’s ai with anthropic’s claude, and there was no character limit to inputs and no login required. With bing, it had a strict character limit, and required that I create an account and login.

      • Varyk@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        3
        arrow-down
        1
        ·
        8 months ago

        I can’t see a huge difference between openai 3.5 and Claude on ddg, but they are both beating copilot hands down.

        If I ask copilot a specific question, like when was the first year this product was released in the United States, it’ll describe the product stuff tell me about the product without mentioning the year, while the ddg chatbots both answer a similar question immediately and accurately.

        My only problem is that unless I’m missing something, ddg botsdon’t provide the source they’re getting the information from, which I do like a lot with copilot, that I can make sure the source they’re pulling the information from is accurate immediately because the link is provided.

        • ShadowCat@lemmy.dbzer0.com
          link
          fedilink
          English
          arrow-up
          2
          ·
          8 months ago

          I’m pretty sure the bots on DDG don’t have internet access, just try asking a question about anything that happened very recently (e.g. last week) and they won’t be able to answer.

  • Kraiden@kbin.run
    link
    fedilink
    arrow-up
    32
    arrow-down
    27
    ·
    8 months ago

    Aaaaand the DDG enshitification begins… continues? continues.

  • exanime@lemmy.today
    link
    fedilink
    English
    arrow-up
    3
    ·
    8 months ago

    I’m finding DDG results very inconsistent as of late.

    I was trying to find an old podcast about the Stanford jail experiment. Couldn’t remember it was Stanford so my prompt was: radiolab jail experiment

    I got a whole page of a Netflix jail experiment show. The entire page were links to Netflix or reviews or news about the Netflix show. Not a single mention of Stanford or radiolab at all

    Switched my prompt to: “radiolab” jail experiment And immediately got what I was looking for.

    Gotta say, not impressed with the results as off late

    • sugar_in_your_tea@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      4
      ·
      8 months ago

      Yeah, DDG seems to have a heavy skew toward newer content. I’ve found myself using a lot of "term" because it’s trying to find something new when I’m looking for something old.

      I still much prefer it to Google, but it could be a lot better.

  • antler@feddit.rocks
    link
    fedilink
    English
    arrow-up
    1
    ·
    edit-2
    8 months ago

    Been helped a lot by Brave Search’s built in AI. Privacy search engines have always been somewhat more unreliable, and with Google searches going to crap now, sometimes that AI answer below the search answers my question or points me in the right direction when non of the results do. If AI’s going to pollution the search results might as well use it to alleviate the mess it made.

    Still would a loved to see DDG go the brave model and actually self host an open model instead of just embedding ChatGPT.

    Edit: nevermind, see they offer both self hosted and ChatGPT3.

  • snownyte@kbin.social
    link
    fedilink
    arrow-up
    13
    arrow-down
    14
    ·
    8 months ago

    First people thought Firefox was the beacon of privacy, but they’re leaning on AI.

    Now DDG is doing it?

    What timeline are we in?

    • BetaDoggo_@lemmy.world
      link
      fedilink
      English
      arrow-up
      22
      arrow-down
      1
      ·
      8 months ago

      All of Firefox’s ai initiatives including translation and chat are completely local. They have no impact on privacy.

    • someguywithacomputer@lemmynsfw.com
      link
      fedilink
      English
      arrow-up
      6
      ·
      8 months ago

      The issue with ai is that you usually have to be logged in with a Gmail account or something before you’re allowed to use it. If DDG is letting people use it anonymously with no login, that’s a hell of a lot better than any alternative besides self hosting your own.

    • TheGrandNagus@lemmy.world
      link
      fedilink
      English
      arrow-up
      6
      arrow-down
      2
      ·
      edit-2
      8 months ago

      Firefox’s AI will be run entirely locally, information you feed it will be completely optional, and allegedly uses ethical training data where nothing was stolen. It’s private.

      It will be used for things like offline translation, finding alternative sources for articles, spotting fake reviews, in the future better text to speech and image recognition for accessibility purposes, etc.

      I think you’re having a kneejerk reaction of “All AI is bad, we must reject technology and embrace tradition”, which isn’t the right response to have. There are valid usecases, IMO, particularly for people with usability requirements.

      Besides, the usual people who complain about Mozilla will complain if Firefox implements AI features, and they’ll also complain if they don’t, saying they’re falling further behind. Mozilla can’t win with you people.

  • ShadowCat@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    2
    arrow-down
    6
    ·
    8 months ago

    this was the perfect opportunity to integrate it with search like bard and ms copilot… but no, they will just offer them just like that, with hallucinations and all. claude 1.2 instant and gpt 3.5 turbo are not even that good

    • wewbull@feddit.uk
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      3
      ·
      8 months ago

      Yes. They should pass on all the things you’ve doing to the data harvesting overlords. Why the hell wouldn’t they?

        • BetaDoggo_@lemmy.world
          link
          fedilink
          English
          arrow-up
          2
          arrow-down
          1
          ·
          8 months ago

          The issue is that they have no way of verifying that. We’d have to trust 2 other companies in addition to DDG.

          • ShadowCat@lemmy.dbzer0.com
            link
            fedilink
            English
            arrow-up
            1
            ·
            8 months ago

            they could’ve made it optional, then if someone wanted to, they could opt in. there are also open source models like llama and mistral which they can run on their own hardware. but that would probably be too expensive for DDG