I was watching the RFK Jr questioning today and when Bernie was talking about healthcare and wages I felt he was the only one who gave a real damn. I also thought “Wow he’s kinda old” so I asked my phone how old he actually was. Gemini however, wouldnt answer a simple, factual question about him. What the hell? (The answer is 83 years old btw, good luck america)

    • halcyoncmdr@lemmy.worldEnglish
      0·
      2 years ago

      To be honest, that seems like it should be the one thing they are reliably good at. It requires just looking up info on their database, with no manipulation.

      Obviously that’s not the case, but that’s just because currently LLMs are a grift to milk billions from corporations by using the buzzwords that corporate middle management relies on to make it seem like they are doing any work. Relying on modern corporate FOMO to get them to buy a terrible product that they absolutely don’t need at exorbitant contract prices just to say they’re using the “latest and greatest” technology.

      • SmoothLiquidation@lemmy.worldEnglish
        0·
        2 years ago

        To be honest, that seems like it should be the one thing they are reliably good at. It requires just looking up info on their database, with no manipulation.

        That’s not how they are designed at all. LLMs are just text predictors. If the user inputs something like “A B C D E F” then the next most likely word would be “G”.

        Companies like OpenAI will try to add context to make things seem smarter, like prime it with the current date so it won’t just respond with some date it was trained on, or look for info on specific people or whatnot, but at its core, they are just really big auto fill text predictors.

      • JohnnyFlapHoleSeed@lemmy.worldBannedEnglish
        0·
        2 years ago

        Yeah, I still struggle to see the appeal of Chatbot LLMs. So it’s like a search engine, but you can’t see it’s sources, and sometimes it ‘hallucinates’ and gives straight up incorrect information. My favorite was a few months ago I was searching Google for why my cat was chewing on plastic. Like halfway through the AI response at the top of the results it started going on a tangent about how your cat may be bored and enjoys to watch you shop, lol

        So basically it makes it easier to get a quick result if you’re not able to quickly and correctly parse through Google results… But the answer you get may be anywhere from zero to a hundred percent correct. And you don’t really get double check the sources without further questioning the chat bot. Oh and LLM AI models have been shown to intentionally lie and mislead when confronted with inaccuracies they’ve given.

    • tal@lemmy.todayEnglish
      0·
      2 years ago

      I mostly can’t understand why people are so into “LLMs as a substitute for Web search”, though there are a bunch of generative AI applications that I do think are great. I eventually realized that for people who want to use their cell phone via voice, LLM queries can be done without hands or eyes getting involved. Web searches cannot.

      • Kairos@lemmy.todayEnglish
        0·
        2 years ago

        Because web search was intentionally hobbled by Google so people are pushed to make more searches and see more ads.

          • Kairos@lemmy.todayEnglish
            0·
            2 years ago

            Do any of them actually work? As in, you search something and it gives you relevant results to the whole thing you typed in?

            • czl@lemmy.dbzer0.comEnglish
              0·
              2 years ago

              I’ve used DuckDuckGo for a long time, so I would say yes. But the best way to figure that out is just to try it for a while. There is literally nothing to lose.

            • chonglibloodsport@lemmy.worldEnglish
              0·
              2 years ago

              It’s always going to depend on what you’re searching for. I just tried searching for home coffee roasting on Swiss Cows and all of the results were legit, no crappy spam sites.

              Marginalia is great for finding obscure sites but many normal sites don’t show up there. Million Short is a similar idea but with a different approach to achieving it.

              The problem of search is actually extremely hard because there are millions of scam and spam sites out there that are full of ads and either AI slop or literally stolen content from other popular sites. Somehow these sites need to be blocked in order to give good results. It’s a never-ending, always-evolving battle, just like blocking spam in email (I still have to check my spam folder all the time because legit emails end up flagged as spam).

          • Nollij@sopuli.xyzEnglish
            0·
            2 years ago

            DDG and Ecosia are proxies for Bing. I didn’t check, but I’m guessing the others are too. Most “independent” search engines are.

            The major exception is Startpage, which is a proxy for Google.

              • Nollij@sopuli.xyzEnglish
                0·
                2 years ago

                That one’s pretty obvious. From their main page:

                Startpage delivers Google search results via our proprietary personal data protection technology.

      • Squorlple@lemmy.worlddeleted by creatorEnglish
        0·
        2 years ago

        Would saying “Gemini, open the Wikipedia page for Bernie Sanders and read me the age it says he is”, for example, suffice as a voice input that both bypasses subject limitations and evades AI bullshitting?

        • brb@sh.itjust.worksEnglish
          0·
          2 years ago

          Gemini refuses to answer

          Copilot seems to know the current date and calculates the age from that

          ChatGPT is clueless

        • DreamlandLividity@lemmy.worldEnglish
          0·
          2 years ago

          Idk if it bypasses limitations, you can try. As for bullshiting, no. The AI almost certainly does not have the ability to go and open a webpage. If it was trained on wikipedia, it may or may not give you the age listed at the time of it’s training. If not, it will likely take a different source and pretend it is from wikipedia. Either way, it will likely bullshit you about doing what you asked while giving you outdated/missourced information.

          Now the number may be correct, I imagine Bernies real age is readily available, but it will confidently lie about how it got the information.

  • Robotunicorn@lemmy.worldEnglish
    0·
    2 years ago

    That’s because Google is not only a shit company, it’s also in Trump’s pocket. They changed the name of the Gulf of Mexico for US users. WTF?

  • Jo Miran@lemmy.mlEnglish
    0·
    2 years ago

    It looks like they disabled responses about current major political figures.

    • BatrickPateman@lemmy.worldEnglish
      0·
      2 years ago

      As OP’s query regarding RFK Jr. yielded a response he can be considered “not a major political figure” then, I guess. Nice burn, Gemini 🤭.

    • Dozzi92@lemmy.worldEnglish
      0·
      2 years ago

      Dunno if your typo was intentional or not, but all I see in this thread is that somehow a typo is a way to bypass whatever block they have on discussion related to political figures. Which is bonkers. The great minds at the Goog somehow missed a pretty obvious workaround.

  • jeeptrap@lemmy.worldEnglish
    0·
    2 years ago

    Google has implemented restrictions on Gemini’s ability to provide responses on political figures and election-related topics. This is primarily due to the potential for AI to generate biased or inaccurate information, especially in the sensitive and rapidly evolving landscape of politics. Here are some of the key reasons behind this decision:

    • Mitigating misinformation: Political discourse is often rife with misinformation and disinformation. AI models, like Gemini, learn from vast amounts of data, which can include biased or inaccurate information. By limiting responses on political figures, Google aims to reduce the risk of Gemini inadvertently spreading misinformation.
    • Avoiding bias: AI models can inherit biases present in their training data. This can lead to skewed or unfair representations of political figures and their viewpoints. Restricting responses in this area helps to minimize the potential for bias in Gemini’s output.
    • Preventing manipulation: AI-generated content can be used to manipulate public opinion or influence elections. By limiting Gemini’s involvement in political discussions, Google hopes to reduce the potential for its technology to be misused in this way.
    • Maintaining neutrality: Google aims to maintain a neutral stance on political matters. By restricting Gemini’s responses on political figures, they can avoid the appearance of taking sides or endorsing particular candidates or ideologies. While these restrictions may limit Gemini’s ability to provide information on political figures, they are in place to ensure responsible use of AI and to protect the integrity of political discourse.
          • geneva_convenience@lemmy.mlEnglish
            0·
            2 years ago

            Does not use the word concentration camp

            Despite the injustices they faced many Japanese sought to demonstrate their loyalty to the Unites States

            Look at my amazing propaganda

            My sides

            • LeeNeighoff@lemmy.worldEnglish
              0·
              2 years ago

              I didn’t say it was perfect or even unbiased, just that it’s better than denying something happened or refusing to talk about it.

                • LeeNeighoff@lemmy.worldEnglish
                  0·
                  2 years ago

                  To each their own, I suppose. I prefer having the info generally available, so I can find various accounts of an event and form my own opinions. If the information is actively suppressed, it’s a lot more difficult to find out what really happened than if the info is available but some sources are (heavily) biased.

  • Jumi@lemmy.worldEnglish
    0·
    2 years ago

    The only mildly infuriating thing is that people keep posting this shit

    • Mr_Dr_Oink@lemmy.worldEnglish
      0·
      2 years ago

      This will be a blanket filter on his name, not his age. And if anything they are trying to de-politicise gemini.

      Not that i agree with that practice, personally.

      But its not fair to say they are politicising age by omitting any responses related to a political figure.

    • SmokeyDope@lemmy.worldEnglish
      0·
      2 years ago

      Local LLM gang represent! ✌️

      Local LLMs aren’t perfect but are getting more usable. There’s abliterated models and uncensored fine tunes to choose from if you don’t like your LLM rejecting your questions.