You know how Google’s new feature called AI Overviews is prone to spitting out wildly incorrect answers to search queries? In one instance, AI Overviews told a user to use glue on pizza to make sure the cheese won’t slide off (pssst…please don’t do this.)

Well, according to an interview at The Vergewith Google CEO Sundar Pichai published earlier this week, just before criticism of the outputs really took off, these “hallucinations” are an “inherent feature” of  AI large language models (LLM), which is what drives AI Overviews, and this feature “is still an unsolved problem.”

  • MNByChoice@midwest.social
    link
    fedilink
    English
    arrow-up
    133
    arrow-down
    1
    ·
    5 months ago

    Good. Nothing will get us through the hype cycle faster than obvious public failure. Then we can get on with productive uses.

      • 9488fcea02a9@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        38
        arrow-down
        16
        ·
        edit-2
        5 months ago

        I hate the AI hype right now, but to say the entire thing should fail is short sighted.

        Imagine people saying the following: “The internet is just hype. I get too much spam emails. I hope the entire thing is a catastrophic failure.”

        Imagine we just shut down the entire internet because the dotcom bubble was full of scams and overhyped…

          • Goodtoknow@lemmy.ca
            link
            fedilink
            English
            arrow-up
            17
            arrow-down
            3
            ·
            5 months ago

            The peak of computer productivity was spreadsheets and smb shares in the '90s everything else has been downhill in terms of increase of distraction and time wasting inefficiencies.

          • slumberlust@lemmy.world
            link
            fedilink
            English
            arrow-up
            1
            ·
            5 months ago

            There is hope! The UK just passed some comprehensive IoT security rules with teeth. An actual win in this megalomaniac capitalists dream of an economy!

        • TurtleJoe@lemmy.world
          link
          fedilink
          English
          arrow-up
          13
          ·
          5 months ago

          The Internet immediately worked, which is one big difference. The dot com financial bubble has nothing to do with the functionality of the internet.

          In this case, there is both a financial bubble, and a “product” that doesn’t really work, and which they can’t make any better (as he admits in this article.)

          It was obvious from day 1 how useful the Internet would be. Email alone was revolutionary. We are still trying to figure out what the real uses for LLM are. There appear to be some valid use cases outside of creating spam and plagiarizing other people’s work, but it doesn’t appear to be any kind of revolutionary technology.

          • 9488fcea02a9@sh.itjust.works
            link
            fedilink
            English
            arrow-up
            6
            arrow-down
            1
            ·
            5 months ago

            “product” that doesn’t really work, and which they can’t make any better

            LLMs “dont work” because people are promising idiotic things and being used recklessly for things they are not good at. This is like saying a chainsaw is a failed product because it’s not good at slicing sushi

            It was obvious from day 1 how useful the Internet would be. Email alone was revolutionary

            Hindsight 20/20. There were a lot of people smarter than you and i predicting that the internet was just a fad

          • nickwitha_k (he/him)@lemmy.sdf.org
            link
            fedilink
            English
            arrow-up
            5
            arrow-down
            1
            ·
            5 months ago

            Summarizing is something that it does very well. Still not 100% but, when using RAG and telling it “don’t make shit up” can result in pretty good compute efficiency and results.

          • brbposting@sh.itjust.works
            link
            fedilink
            English
            arrow-up
            4
            arrow-down
            1
            ·
            5 months ago

            There appear to be some valid use cases outside of creating spam and plagiarizing other people’s work

            Like translation, which has already taken money out of the pockets of 40% of translators?

            + customer service, incl. sources

            November 2022: ChatGPT is released

            April 2024 survey: 40% of translators have lost income to generative AI - The Guardian

            Also of note from the podcast Hard Fork:

            There’s a client you would fire… if copywriting jobs weren’t harder to come by these days as well.

            Customer service impact, last October:

            And this past February - potential 700 employee impact at a single company:

            If you’re technical, the tech isn’t as interesting [yet]:

            Overall, costs down, capabilities up (neat demos):

            Hope everyone reading this keeps up their skillsets and fights for Universal Basic Income for the rest of humanity :)

        • utopiah@lemmy.world
          link
          fedilink
          English
          arrow-up
          6
          arrow-down
          2
          ·
          5 months ago

          Genuinely curious, what pieces do you suggest we can keep from LLM/GenAI/etc?

          • KevonLooney@lemm.ee
            link
            fedilink
            English
            arrow-up
            12
            arrow-down
            3
            ·
            5 months ago

            ?

            Have you never used any of these tools? They’re excellent at doing simple things very fast. But it’s like a word processor in the 90s. It’s just a tool, not the font of all knowledge.

            I guess younger people won’t know this, but word processor programs were very impressive when they first came out. They replaced typewriters; a page printed from a printer looked much more professional than even the best typewriters. This lent an air of credibility to anything that was printed from a computer because it was new and expensive.

            Think about that now. Do you automatically trust anything that’s just printed on a piece of paper? No, because that’s stupid. Anyone can just print whatever they want. LLMs are like that now. They can just say whatever they want. It’s up to you to make sure it’s true.

          • Microw@lemm.ee
            link
            fedilink
            English
            arrow-up
            4
            arrow-down
            1
            ·
            5 months ago

            The main field where they are already actively in professuonal use are rough drafts in creative fields: quickly generate possible outlines for a text, a speech, an art piece. Visualize where something could be going, in order to decide which direction to pick.

            Also, models that work differently from the GPTs are already in use in science, scanning through huge amounts of texts in archives to help analyzing or search for something in particular. Help find patterns in things for studies. Etc.

            The “personal assistant AI” thing obviously isnt quite working yet. I think it will take some time and models with a different technological structure (not GPT) to achieve progress in that regard.

          • hglman@lemmy.ml
            link
            fedilink
            English
            arrow-up
            6
            arrow-down
            6
            ·
            5 months ago

            Using it to generate things that you double check. Transforming generative work to review work is a boost in productivity. So writing of any kind, art, etc. asking the llm for facts without context is a gross mistake. Prompting it to generate a specific paragraph in a larger, technical or regulator document is useful.