from Hacker News

Meta’s live demo fails; “AI” recording plays before the actor takes the steps

by personjerry on 9/18/25, 8:50 PM with 342 comments

  • by alangibson on 9/18/25, 10:11 PM

    That wasn't prerecorded, but it was rigged. They probably practiced a few times and it confused the AI. Still it's no excuse. They've dropped Apollo-program level money on this and it's still dumb as a rock.

    I'm endless amazed that Meta has a ~2T market cap, yet they can't build products.

  • by AdmiralAsshat on 9/18/25, 10:19 PM

    The Kotaku article on this had a really nice final zinger[0]:

    > Oh, and here’s Jack Mancuso making a Korean-inspired steak sauce in 2023.

    > https://www.instagram.com/reel/Cn248pLDoZY/?utm_source=ig_em...

    0: https://kotaku.com/meta-ai-mark-zuckerberg-korean-steak-sauc...

  • by patrickhogan1 on 9/18/25, 10:36 PM

    Credit where it’s due: doing live demos is hard. Yesterday didn’t feel staged—it looked like the classic “last-minute tweak, unexpected break.” Most builders have been there. I certainly have (I once spent 6 hours at a hackathon and broke the Flask server keying in a last minute change on the steps of the stage before going on).

  • by sampton on 9/18/25, 10:08 PM

    I bet they rehearsed a dozen times and never failed as bad live. Got to give them props for keeping the live demos. Apple has neutered its demos so much it's now basically 2 hr long commercials.

  • by yallpendantools on 9/18/25, 10:48 PM

    I have a friend who does magic shows. He sells his shows as magic and stand-up comedy. It's both live entertainment, okay, but he is the only person I've ever seen use that tagline. We went to see him perform once and everything became clear when he opened the night.

    "This is supposed to be a magic show," he told us. "But if my tricks fail you can laugh at it and we'll just do stand-up comedy."

    Zuck, for a modest and totally-reasonable fee, I will introduce you to my friend. You can add his tricks (wink wink) to your newly-assembled repertoire of human charisma.

  • by skhameneh on 9/18/25, 9:57 PM

    As much as it'll be "interesting" to see how models behave in real world examples (presumably similarly to how the demos went), I'm not convinced this is a premade recording like what seems to be implied.

    I'm imagining this is an incomplete flow within a software prototype that may have jumped steps and lacks sufficient multi-modal capability to correct.

    It could also be staged recordings. But, I don't think it really matters. Models are easily capable of working with the setup and flow they have for the demo. It's real world accuracy, latency, convenience, and other factors that will impact actual users the most.

    What's the reliability and latency needed for these to be a useful tool?

    For example, I can't imagine many people wanting to use the gesture writing tools for most messages. It's cool, I like that it was developed, but I doubt it'll see substantial adoption with what's currently being pitched.

  • by Jackson__ on 9/19/25, 1:27 PM

    This appears to be a classic vision fail on the VLM's part. Which is entirely unsurprising for anyone who has used open VLMs for anything except ""benchmarks"" in the past two god damn years. The field is in a truly embarrassing state, where they pride themselves how it can solve equations off a blackboard, yet couldn't even accurately read a d20 dice roll among many other things. I've tried (and failed) to have VLMs accurately caption images for such a long time, yet anytime I check on the output it is blindingly clear that these models are awful at actually _seeing things_.

  • by throwaway13337 on 9/19/25, 12:08 AM

    I don't think it was rigged.

    Having claude run the browser and then take a screenshot to debug gives similar results. It's why doing so is useless even though it would be so very nice if it worked.

    Somewhere in the pipeline, they get lazy or ahead of themselves and just interpret what they want to in the picture they see. They want to interpet something working and complete.

    I can imagine it's related the same issue with LLMs pretending tests work when they don't. They're RL trained for a goal state and sometimes pretending they reached the goal works.

    It wasn't the wifi - just genAI doing what it does.

  • by kiratp on 9/19/25, 12:41 AM

    So much negativity.

    I’m just excited that our industry is lead by optimists and our culture enables our corporations to invest huge sums into taking us forward technologically.

    Meta could have just done a stock buyback but instead they made a computer that can talk, see, solve problems and paint virtual things into the real world in front of your eyes!

    I commend them on attempting a live demo.

  • by ryandrake on 9/18/25, 9:51 PM

    I love how they randomly blame the WiFi network, like anyone is going to buy it.

  • by ModernMech on 9/19/25, 3:12 PM

    This is how you know AI will not live up to the hype. You have the highest paid people, at one of the richest companies in the world, building the AI and with unlimited access to the best models and the most talent.

    And they still can't pull off a keynote.

    So then... what does AI have to offer me? Because I would have thought, as Sam Altman put it, having an expert PhD level researcher in all subjects in my pocket could maybe help me pull off a tech demo. But if it can't help them, the people who actually made the thing, on their very high stakes public address where everything is on the line, then what's it supposed to do for the rest of us in our daily lives?

    Because it seems more and more, AI is a tool that helps you stage your own very public humiliation.

  • by blinding-streak on 9/18/25, 11:22 PM

    Everything is always so cringe with Facebook and Zuck. It was always doomed to fail.

  • by watersb on 9/18/25, 10:10 PM

    Never work with children, animals, and puppets.

    https://tvtropes.org/pmwiki/pmwiki.php/Main/NeverWorkWithChi...

  • by reader9274 on 9/18/25, 9:58 PM

    Should've downloaded more ram for the wifi to work better

  • by pera on 9/18/25, 9:29 PM

    There is a second part that is equally bad, but with Zuck:

    https://old.reddit.com/r/interestingasfuck/comments/1nkbqyk/...

  • by MattDaEskimo on 9/18/25, 10:16 PM

    More than likely the full response was kept as context despite being interrupted.

    Notably though, the AI was clearly not utilizing its visual feed to work alongside him as implied

  • by eviks on 9/19/25, 2:47 AM

    It's not a fail, but the correct demonstration of the quality of the service

  • by hu3 on 9/18/25, 10:54 PM

    I hope they keep doing live demos. This is much better than prerecorded videos.

  • by jjbinx007 on 9/18/25, 9:35 PM

    This is like a Black Mirror episode. Also, is it a conscious decision to make the TTS sound so robotic?

  • by pciexpgpu on 9/18/25, 10:56 PM

    It's an ad network with an attached optional pair of glasses.

    It's the platform Zuck always wanted to own but never had the vision beyond 'it's an ad platform with some consumer stuff in it'.

    I am super impressed with the hardware (especially the neural band) but it just so happens that a very pricey car is being directly sold by an oil company as a trojan horse.

    We all know what the car is for unfortunately.

    I can't wait to see what Apple has in store now in terms of the hardware.

  • by duxup on 9/19/25, 1:35 PM

    I'm sure there were engineers who were "yo guies don't do this it's not ready to show" and ... naw the yes men "won".

  • by rkagerer on 9/19/25, 1:30 AM

    Why the bleep do they still rely on wifi at conferences like this?? I always insist on a wired connection on its own, dedicated, presenter vlan. Is this running on wifi-only glasses or something? Is that the only medium they can present the tech on? Could they have shielded the room the guy's in?

  • by justinator on 9/19/25, 2:00 AM

    Deja Vu on this demo -- I've seen this one before,

    https://www.youtube.com/watch?v=TYsulVXpgYg

  • by JCM9 on 9/19/25, 12:20 AM

    This is like something right out of the show Silicon Valley. You couldn’t have scripted a more cringe-worth demo.

    It’s like they mashed up the AI and metaverse into a dumpster fire of aimless tech product gobodlygook. The AI bubble can’t pop soon enough so we can all just get back to normal programming.

  • by racl101 on 9/19/25, 5:17 PM

    The main takeaway for me was: do not interrupt AI while it's giving exposition. No matter how trivial it may seem. It will throw it off-kilter.

  • by BoredPositron on 9/19/25, 6:49 AM

    Why is Meta not able to produce one proper keynote in their 21 year history? At this point it must be on purpose just to be relatable or something.

  • by evanjrowley on 9/19/25, 1:33 AM

    This does not deter me from possibly buying one. The concept is pretty cool and appealing to those who want a distraction free lifestyle. Even if there's a screen in front of you at all times, at least you won't need to hold something in your hands to be able to operate it. That alone is a significant win.

  • by maxlin on 9/19/25, 2:14 AM

    Give them a break, small indie company

  • by liendolucas on 9/18/25, 11:36 PM

    LMAO. Billions of dollars for this, seriously? Makes Bill Gates Win95 presentation BSOD fail look professional.

    And LMAO for all the companies out there burning money for getting on the train of AI just because everyone does so.

  • by callbacked on 9/19/25, 12:00 AM

    These demos whether good or bad go in meta's favor I think

    Successful demo? sweet! people will rave about it for a bit

    Catastrophic failure? sweet! people will still talk about it and for even longer now!

  • by jayd16 on 9/19/25, 2:34 AM

    I do wonder what it will take to replace the default voice command handler. Will it be as locked down as possible or will Meta be happy with any adoption at all?

  • by paxys on 9/19/25, 12:38 AM

    How exactly is this "AI recording plays before the actor takes the steps"?

    This place really is Reddit these days, so I guess the link is apt.

  • by jayd16 on 9/19/25, 12:00 AM

    What part is supposed to be prerecorded?

  • by phodo on 9/19/25, 3:34 AM

  • by yieldcrv on 9/19/25, 2:38 AM

    The interpretation of this is pretty dumb since its not a recording

  • by self_awareness on 9/18/25, 10:09 PM

    Those WiFi's man, they're always trouble

  • by kovac on 9/19/25, 3:44 AM

    Forgot to reset the context after a test run? :p

  • by rsynnott on 9/19/25, 12:08 PM

    Thing is, even if this had _worked_, it would still have been, ah, a bit shit. Absolutely farcical.

  • by nba456_ on 9/18/25, 10:17 PM

    still better than the pre-taped apple events. happy to see these products in action

  • by josefritzishere on 9/18/25, 9:35 PM

    AI is hot trash. When will this river of garbage stop?

  • by Teleoflexuous on 9/19/25, 12:42 AM

    There's a simple explanation that isn't 'prerecorded'. I'd be very happy to accuse Meta of faking a demo, but that's 1) just a weird way to fake a demo and 2) effect that has easier explanation.

    You ask AI how to do something. AI generates steps to do that thing. It has concept of steps, so that when you go 'back' it goes back to the last step. As you ask how to do something, it finishes explaining general idea and goes to first step. You interrupt it. It assumes it went through the first step and won't let you go back.

    The first step here was mixing some sauces. That's it. It's a dumb way to make a tool, but if I wanted to make one that will work for a demo, I'd do that. Have you ever tried any voice thing to guide you through something? Convincing Gemini that something it described didn't happen takes a direct explanation of 'X didn't happen' and doesn't work perfectly.

    It still didn't work, it absolutely wasn't wi-fi issue and lmao, technology of the future in $2T company, it just doesn't seem rigged.

  • by anigbrowl on 9/19/25, 7:45 AM

    Legs! Part II

  • by jwpapi on 9/19/25, 1:03 AM

    Maybe it’s the new Silicon Valley trend to fuck up demos so people talk about it more?

    You know there is no such things as bad publiciity..

  • by steveBK123 on 9/19/25, 11:24 PM

    So much FOMO, hopium and fraud in the space

    We'll be talking about how obvious it all was 20 years from now

  • by yodsanklai on 9/19/25, 12:01 AM

    Yet Meta stock is at all time high

  • by johnfn on 9/19/25, 1:13 AM

    The mocking, gleeful negativity here concerns me. I am worried that with some of these more polarized topics that the discussions on HN are becoming closer to those on Reddit. The fact that the highest upvoted post on this thread is just a link to Reddit isn't doing much to help me feel better. And I've been here for at least a decade, so I don't think this is the noob illusion.

    I have no illusions about Zuckerberg. He's done some pretty bad stuff for humanity. But I think AI is pretty cool, and I'm glad he's pushing it forward, despite mishaps. People don't have to be black or white, and just because they did something bad in one domain doesn't make everything they touch permanently awful.

  • by timpera on 9/18/25, 9:42 PM

    The failures on stage were kind of endearing, to be honest, especially the one with Zuck. Plus the products seem really cool, I hope I'll be able to try them out soon.

  • by nmilo on 9/19/25, 1:26 AM

    It's well known Meta AI is shit. But I could probably make an app that can run this demo in an afternoon. The glasses part here is insane and I don't know why everyone is fixated on the tacky AI part. It's like if I invented the car and you complained that it's really hard to crank the windows down. Be happy it's even there!

  • by lifthrasiir on 9/18/25, 11:34 PM

    One important thing to note: demo didn't fail! (Or, at least not in the way people usually think of)

    > You've already combined the base ingredients, so now grate a pear to add to the sauce.

    This is actually the correct Korean recipe for bulgogi steak sauce. The only missing piece here is that the pear has to be Pyrus pyrifolia [1], not the usual pear. In fact every single Korean watching the demo was complaining about this...

    [1] https://en.wikipedia.org/wiki/Pyrus_pyrifolia