• taxon@lemmy.world
    link
    fedilink
    English
    arrow-up
    18
    ·
    17 hours ago

    A classic example of realworld AI implementation. I wonder how many times they rehearsed the cooking scene before deciding it was never going to work.

  • I Cast Fist@programming.dev
    link
    fedilink
    arrow-up
    8
    ·
    16 hours ago

    I see Mark Cuckberg has become an adept of the “BSoD in front of everyone”, now that Microsoft no longer does that. Remember that metaverse marketing?

  • brachiosaurus@mander.xyz
    link
    fedilink
    arrow-up
    33
    arrow-down
    3
    ·
    23 hours ago

    You are all falling for meta advertising, if it wasn’t for this mishap nobody would know they are releasing a new product

  • Denjin@feddit.uk
    link
    fedilink
    arrow-up
    93
    ·
    1 day ago

    Even if this worked perfectly, ignoring the fact that it’s clearly setup for the camera to recognise certain things and is in no way a genuine demonstration, what is the point of this? By the time it’s even responded to his first “Hey meta” he could have typed “korean steak sauce recipe” into his search engine of choice and got back several dozen decent results in seconds.

    What is the problem that these LLMs and chatbots are the solution for? It’s like they’re all desperately trying to market some fancy new type of barely functioning legs to everyone when we already have legs, and arms, and cars, and bicycles.

    • someacnt@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      1
      ·
      6 hours ago

      To be fair, it is helpful to do grunt repetitive work which is only slightly different from the established formula. For me, it solved a quick math olympiad-esque problem which appeared in research.

      But it is never conscious or creative.

    • voodooattack@lemmy.world
      link
      fedilink
      arrow-up
      35
      arrow-down
      1
      ·
      1 day ago

      That’s not the point. He’s planning to harvest data about the environment in your home (what products you have around you, which brands do you prefer, etc) for better ad targeting and whatnot.

      I guarantee that a lot of people will use it; and not because it does a great job, but out of curiosity, peer pressure, or abject laziness.

      • Coyote_sly@lemmy.world
        link
        fedilink
        English
        arrow-up
        4
        ·
        17 hours ago

        Well sure, but you can’t harvest data using your one legged mech suit of everyone keeps riding around on their perfectly functional bike.

        LLMs just don’t have a real use case for most people. That’s the core issue here, and it isn’t one that’s going to get solved anytime soon.

        • voodooattack@lemmy.world
          link
          fedilink
          arrow-up
          4
          ·
          edit-2
          17 hours ago

          I’m pretty sure lots of people use Google Home Assistant or Amazon’s Alexa. Now slap a camera on it and hook it to an LLM and you’re all set. That’s what meta is about to do.

          • JcbAzPx@lemmy.world
            link
            fedilink
            English
            arrow-up
            1
            ·
            16 hours ago

            Meta isn’t really a widget seller, though. Alexa and the Google thing only have the user base they do because they jam it into every device they sell.

            • voodooattack@lemmy.world
              link
              fedilink
              arrow-up
              2
              ·
              16 hours ago

              It’s not a huge leap for them though. They could do it. They already sell VR gear that’s relatively popular, and they have the resources and reach to market anything. If it allows them to spy on people to this level, watch them subsidise any device they sell to hell and back.

              Even acquiring a company that already sells consumer electronics/appliances and subsidising their products isn’t impossible for them. Like I said, they have the resources.

      • ideonek@piefed.social
        link
        fedilink
        English
        arrow-up
        2
        ·
        16 hours ago

        Which will still not be enught to make ads work. It’s so much effort and abuse for nothing.

    • Alcoholicorn@mander.xyz
      link
      fedilink
      arrow-up
      3
      arrow-down
      2
      ·
      1 day ago

      The idea is the AI would automatically look at what you have and come up with something, substituting ingredients for what you have as necessary.

      “korean steak sauce recipe” into his search engine of choice and got back several dozen decent results in seconds.

      Have you tried looking for a recipe in the last 10 years or so? 10 pages of fluff with a recipe the author cobbled together from other recipes and guess work and made exactly 1 time if that at the bottom.

      • greybeard@feddit.online
        link
        fedilink
        English
        arrow-up
        9
        arrow-down
        1
        ·
        20 hours ago

        You know what would be more useful than this AI and wouldn’t cost billions of dollars? If Facebook made a simple recipe that didn’t have all that fluff. Instead of having AI try to come up with it in the fly, you could just have premade recipes. Wouldn’t that be grand. Oh wait, that wouldn’t give the opportunity for the AI to recommend a specific brand of soy sauce or that you buy your spices from Doordash with a 10% discount coupon!

        • MiddleAgesModem@lemmy.world
          link
          fedilink
          arrow-up
          1
          arrow-down
          5
          ·
          17 hours ago

          chatGPT as NEVR suggested a specific bran to me. What is it about anti-AI crusading do you think excuses your deliberate misinformation?

          • greybeard@feddit.online
            link
            fedilink
            English
            arrow-up
            5
            ·
            13 hours ago

            Why do you think advertising companies like Google and Meta are dumping so much money into AI? It isn’t to make a better product for the user.

      • Denjin@feddit.uk
        link
        fedilink
        arrow-up
        10
        arrow-down
        1
        ·
        24 hours ago

        The idea is the AI would automatically look at what you have and come up with something, substituting ingredients for what you have as necessary.

        But as in the staged example they’ve put together here, that required you to find and lay out all your ingredients already so you’ve already done 90% of the work. Are my AI glasses going to be able to scan all my cupboards and fridge and pantry for things first and then go from there? It’s a bad solution to a problem that doesn’t really exist.

        Have you tried looking for a recipe in the last 10 years or so? 10 pages of fluff with a recipe the author cobbled together from other recipes and guess work and made exactly 1 time if that at the bottom.

        Yes, all the time, I even looked at recipes for “Korean style steak sauce” to prove my point and got probably a dozen decent ones straight away, including a couple from very well recognised sites like BBC Good Food.

        And where exactly do you suppose the LLM has scraped together all it’s information from for what you can substitute canola oil or sesame seeds for? Those exact recipes you could just have searched for.

        • MiddleAgesModem@lemmy.world
          link
          fedilink
          arrow-up
          1
          arrow-down
          2
          ·
          17 hours ago

          And where exactly do you suppose the LLM has scraped together all it’s information from for what you can substitute canola oil or sesame seeds for? Those exact recipes you could just have searched for.

          Who cares? It’s still more convenient.

      • MiddleAgesModem@lemmy.world
        link
        fedilink
        arrow-up
        2
        arrow-down
        5
        ·
        17 hours ago

        Cooking is my favorite use of AI. It has completely changed the way I eat, which is much healthier. I also try new things I never would have before. I load up on healthy ingredients, spices and sauces and let it do it’s thing. You can always have it mix it up, or focus on an ingredient that’s more perishable, or focus on something that may be expiring soon. The variability is endless. I’ve been cooking every day for months, never once made the same thing twice.

  • INeedMana@piefed.zip
    link
    fedilink
    English
    arrow-up
    38
    ·
    1 day ago

    I wonder if botched demos are not starting to become intentional. In the end, the more people talk about something, the more successful that marketing action is, right?

    • ideonek@piefed.social
      link
      fedilink
      English
      arrow-up
      14
      ·
      16 hours ago

      I hate it. There is no dumb enought thing, that billioners can do to prove that are truly dumb, without people giving them benefits of playing 4-dimensional chess. The king truly is naked.

      • postmateDumbass@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        9 hours ago

        No sense in spending the extra to get it perfect when your asking them to invest in the concept and the future.

        Also, would set expectations for the next demo to actually work too.

  • SoupBrick@pawb.social
    link
    fedilink
    arrow-up
    124
    arrow-down
    1
    ·
    1 day ago

    “I think the WIFI might be messed up.”

    Bro, I think the AI might be messed up. You don’t get instructions to grate a pear when a device cannot connect to the internet.

  • ɯᴉuoʇuɐ@lemmy.dbzer0.com
    link
    fedilink
    arrow-up
    96
    ·
    1 day ago

    What surprises me isn’t that AI failed, what surprises me is that Zuckerberg believes in his own hyping of AI so much he had the confidence to try to do this at all, live, unstaged. Is this courage, honesty, stupidity, hubris?

    The most favourable explanation I can think of is that they tried it out ahead of the presentation, it worked well enough, so they trusted it could be repeated.

    • njm1314@lemmy.world
      link
      fedilink
      arrow-up
      31
      ·
      1 day ago

      I just don’t know how many times these billionaires have to prove that deep down they’re just dumb asses at heart until people get the message

      • tigeruppercut@lemmy.zip
        link
        fedilink
        arrow-up
        1
        arrow-down
        1
        ·
        6 hours ago

        I mean, elon and zuck aren’t dummies, but unless (or perhaps even if) you’re a literal genius you need other people to constantly be checking your ideas. When you only have sycophants you’re gonna start churning out stupid ideas.

      • Tollana1234567@lemmy.today
        link
        fedilink
        arrow-up
        15
        ·
        1 day ago

        they surround themselves with sycophants thier whole lives, to avoid the uncomfortable possibility that someone says they are wrong.

  • rustydrd@sh.itjust.works
    link
    fedilink
    arrow-up
    22
    ·
    1 day ago

    You’d think they script this and test it a couple of times before showing it to the press. If they can’t manage to do that, maybe the hundreds of billions of investments into their AI program are a teeny tiny bit misplaced.

  • MangoPenguin@lemmy.blahaj.zone
    link
    fedilink
    English
    arrow-up
    73
    ·
    1 day ago

    God that’s so awkward, the super long pauses before it can even start replying, the horrible robot voice it has, and them both trying to pretend it’s a wifi issue lmao.

      • MiddleAgesModem@lemmy.world
        link
        fedilink
        arrow-up
        2
        arrow-down
        2
        ·
        17 hours ago

        What about it? I don’t know the quality of Facebook’s but what I use works extremely well for regular cooking.

    • Tilgare@lemmy.world
      link
      fedilink
      arrow-up
      3
      arrow-down
      25
      ·
      1 day ago

      To be fair (not that Zuck has ever done anything to earn it), wifi at enormous events like that can be super flaky and slow. Thousands of people hitting it simultaneously and it slows to a crawl.

      • Anivia@feddit.org
        link
        fedilink
        arrow-up
        55
        ·
        1 day ago

        Wifi being slow doesn’t make an AI reply bullshit, at worst it makes it reply slower or not at all

      • IninewCrow@lemmy.ca
        link
        fedilink
        English
        arrow-up
        24
        ·
        1 day ago

        These are managed and controlled events. They would have setup dedicated wifi systems to eliminate as much interference as possible. I’ve worked at a couple of small scale public events for companies … they manage things down to the smallest details to try to make sure things work as intended. The bigger the company, the bigger the budget and the more they try to control everything.

        For a company like Facebook hosting a public event like this, they would have controlled absolutely everything, including the wifi … and the only variable that they couldn’t control was the AI and what it would, or not do.

  • nonentity@sh.itjust.works
    link
    fedilink
    arrow-up
    22
    arrow-down
    1
    ·
    1 day ago

    Ridiculing slopvangelical LLM thumpers is the only valid response.

    No one who truly understands how computing works can legitimately argue what played out in this video was the result of ‘bad Wi-Fi’.