

Glitches Humiliated Zuck in Smart Glasses Launch. Meta CTO Explains What Happened (techcrunch.com) 77
When Meta finally unveiled its newest smart glasses, CEO Mark Zuckerberg "drew more snickers than applause," wrote the New York Times. (Mashable points out a video call failing onstage followed by an unsuccessful recipe demonstration.)
Meta chief technology officer Andrew Bosworth later explained the funny reason their demo didn't work, reports TechCrunch, while answering questions on Instagram: "When the chef said, 'Hey, Meta, start Live AI,' it started every single Ray-Ban Meta's Live AI in the building. And there were a lot of people in that building," Bosworth explained. "That obviously didn't happen in rehearsal; we didn't have as many things," he said, referring to the number of glasses that were triggered... The second part of the failure had to do with how Meta had chosen to route the Live AI traffic to its development server to isolate it during the demo. But when it did so, it did this for everyone in the building on the access points, which included all the headsets. "So we DDoS'd ourselves, basically, with that demo," Bosworth added... Meta's dev server wasn't set up to handle the flood of traffic from the other glasses in the building — Meta was only planning for it to handle the demos alone.
The issue with the failed WhatsApp call, on the other hand, was the result of a new bug. The smart glasses' display had gone to sleep at the exact moment the call came in, Bosworth said. When Zuckerberg woke the display back up, it didn't show the answer notification to him. The CTO said this was a "race condition" bug... "We've never run into that bug before," Bosworth noted. "That's the first time we'd ever seen it. It's fixed now, and that's a terrible, terrible place for that bug to show up." He stressed that, of course, Meta knows how to handle video calls, and the company was "bummed" about the bug showing up here... "It really was just a demo fail and not, like, a product failure," he said.
Thanks to Slashdot reader fjo3 for sharing the news.
Meta chief technology officer Andrew Bosworth later explained the funny reason their demo didn't work, reports TechCrunch, while answering questions on Instagram: "When the chef said, 'Hey, Meta, start Live AI,' it started every single Ray-Ban Meta's Live AI in the building. And there were a lot of people in that building," Bosworth explained. "That obviously didn't happen in rehearsal; we didn't have as many things," he said, referring to the number of glasses that were triggered... The second part of the failure had to do with how Meta had chosen to route the Live AI traffic to its development server to isolate it during the demo. But when it did so, it did this for everyone in the building on the access points, which included all the headsets. "So we DDoS'd ourselves, basically, with that demo," Bosworth added... Meta's dev server wasn't set up to handle the flood of traffic from the other glasses in the building — Meta was only planning for it to handle the demos alone.
The issue with the failed WhatsApp call, on the other hand, was the result of a new bug. The smart glasses' display had gone to sleep at the exact moment the call came in, Bosworth said. When Zuckerberg woke the display back up, it didn't show the answer notification to him. The CTO said this was a "race condition" bug... "We've never run into that bug before," Bosworth noted. "That's the first time we'd ever seen it. It's fixed now, and that's a terrible, terrible place for that bug to show up." He stressed that, of course, Meta knows how to handle video calls, and the company was "bummed" about the bug showing up here... "It really was just a demo fail and not, like, a product failure," he said.
Thanks to Slashdot reader fjo3 for sharing the news.
Demo failure not a product failure (Score:5, Interesting)
You can't fail a demo due to a bug in the product if the product is working perfectly. He should accept the failure with more grace ala BSOD on Windows 95 when Bill Gates plugged a plug and pray printer in, rather than gaslighting everyone.
Re: (Score:3)
Thanks for chiming in, Mr. Joyce.
Both a Demo and Product failure. (Score:2)
You can't fail a demo due to a bug in the product if the product is working perfectly. He should accept the failure with more grace ala BSOD on Windows 95 when Bill Gates plugged a plug and pray printer in, rather than gaslighting everyone.
To be fair, Bill probably had no idea he even needed to bring a bible that day. Much less pray.
Had to “hack” and isolate the product from the live AI to simply demo it. While also failing to isolate the demo from the crowd. It wasn’t a demo or product failure. It was both. Isolating your new smart product from the very AI it will rely on for every consumer, screams you’re not ready.
Enough of the childish premature half-ass AI demos for shits and stock price sake. The real worl
Re: (Score:2)
Of you could do what Jobs did and find a way through the demo that won't lead to problems. It's well known that the iPhone OS was highly unstable in the early days and there were lots of bugs ranging from crashes to things like Wi-Fi not working.
What Jobs did was basically spend a month going through and finding a path to demo what he wanted to show while avoiding all the bugs and traps. Sure the bugs were being fixed all the time to get it ready for release 6 months later, but for the demo there had to be
Re: Both a Demo and Product failure. (Score:1)
Re: (Score:2)
The BSOD was caused by the demo using a different scanner than the one that had been tested (and that demo scanner was one that apparently drew more power than it said it would): https://devblogs.microsoft.com... [microsoft.com]
How ironic a CEO lied about the specifications of their product that caused another CEO to not be able to lie as well about theirs.
I envision those two seeing each other at a $30,000 dinner party donating to tax relief efforts for billionaires, and simply nodding to each other as if to say Touche my friend. Tou-fuckin-che.
Re: (Score:2)
What Jobs did was basically spend a month going through and finding a path to demo what he wanted to show while avoiding all the bugs and traps.
Thank you for detailing how the former Apple CEO that was fired from the very company he helped birth, returned to discover the fine art of bullshitting both consumers and investors for shits and stock price sake. When it takes more effort to bullshit a demo than to just give the damn thing, ‘nuff said.
The end result of exactly my point, has been consumers silently being forced to become the beta testers of shit products blatantly lied about to consumers and investors that are absolutely not ready fo
Re:Demo failure not a product failure (Score:5, Insightful)
After watching the videos, I think the demo succeeded in showing that the technology isn't ready. This isn't a BSOD moment.
The demo of the recipe made clear that the AI didn't understand the concept of "first". The video call demonstration showed that the control mechanisms are flaky, not that Zuck just had bad luck on stage. It also showed how intrusive the technology would be. Can you imagine every WhatsApp message popping up in front of you, regardless of what you're doing? Watch the video, you'll be able to imagine.
If you think AI is actually "smart" you're going to be in for a letdown. And I say that as someone who thinks AI *is* a great tool that is useful for many things.
Re: (Score:2)
Re: (Score:2)
Can you imagine every WhatsApp message popping up in front of you, regardless of what you're doing?
If that can happen, then it will not be long before a continuous stream of advertisements are flowing across your vision.
Demo Effect (Score:2)
Re:Demo Effect (Score:5, Funny)
Re: (Score:2)
Apparently reality distortion field beats demo effect.
Re: Demo Effect (Score:2)
Re: (Score:3)
Jobs had it a few times, but he usually just had a backup device ready to switch over to. I recall it at least once with a Mac, and famously with I think an iPhone where be blamed the large number of WiFi devices in the room.
Re: Demo Effect (Score:2)
Re: Demo Effect (Score:3)
Re: (Score:2)
"its", "you're".. *cries in shame*
Don't sweat it - that kind of thing happens to the best of us. Too frequently, I can't even blame my own errors on the "typo" cliche - it's just inattention and a failure to proof-read properly.
Good on you for catching and correcting.
Re: (Score:2)
Jobs was quite relentless, "fix it or your fired" he said, or something to that effect.
Hm, quality management skills right there. Glad I never worked for him.
Re:Demo Effect (Score:4, Informative)
It isn't magic what Steve Jobs did. Nobody is immune to Murphy's Law. "If anything can go wrong, it will"... the key is to reducing what can go wrong. Preparation is the key. Steve Jobs rehearsed for days in advance of major demos. In the case of the iPhone demo, he knew exactly what buttons to press in what order to avoid known bugs.
(The original link might not work, but someone copied the text of the article)
https://www.reddit.com/r/iphone/comments/97jw9f/heres_the_story_of_how_steve_jobs_brilliantly/
He almost certainly did this for every major demonstration. My bet in Zuckerberg learned a lesson, but it does require patience and discipline to do as much work as Jobs did - and most people don't have that.
Re: Demo Effect (Score:2)
did'nt take questions till the end of the demo, and stuck to the script. It's a performance to build client confidence, not a finished product.
Re: (Score:2)
Re: (Score:3)
Jobs had a charisma and our collective impression of him as an innovator even if we knew he was a bit of an ass, deserved or not.
Zuck's time in the public consciousness has been one of a real goober. Didn't help that he was real young when that happened compared to Jobs but his own statements and public presence has not helped either.
Jobs vs Zuckerberg is a real study on how to handle public image.
Re: (Score:2)
No, I don't think this was just the "demo effect."
The BSOD experienced by Bill Gates on screen, that was clearly the "demo effect." It was a truly unexpected interruption to the demo.
If you watch the videos linked in the summary, you can see that the core functionality isn't close to ready. Never mind AI, he did successfully "click" the "answer call" button on the screen (after the first time failed), and it still didn't actually...answer the call.
As someone who has given many, many software demos, complete
Re: (Score:2)
The device had a detachable keyboard. The solution to the random lockups was addi
Re: (Score:2)
I had a similar case in a networks class where we'd designed a server and client with sliding window functionality and some other low level functionality. It worked pretty well but would sometimes lock up and we hadn't had time to look into why.
During the final demo in front of the professor it was going well enough that he saw it was working, then right as he was asking some questions about it, it froze up. We got through the questions and then I suggested we show something in the startup again. If the
Re: (Score:2)
You have no idea how Jobs was immune to it? The answer is they faked it. Here's some random link I just found: https://andrewzuo.com/apples-f... [andrewzuo.com]
At least they tried to do it live (Score:2)
Re: (Score:1)
Did the robots at least have legs?
Of course... (Score:2)
We have only his word that this is actually what happened. And, this being Facebook / Meta, they are known for lying through their teeth.
I wouldn't put it past Zuck to demand his minions make up some explanation that didn't boil down to "our tech sucks".
Re: (Score:2)
The 'explanation' is that the demo triggered all the devices within earshot because apparently a device designed to perform possibly-sensitive actions on your behalf was assigned a model line wide, public audio trigger in order to make it feel more 'natural' or something; rather than some prosaic but functional solution like a trigger button/capacitive touch point/whatever; and that the device just silently fails stupid, no even infor
Re: (Score:2)
The 'explanation' is that the demo triggered all the devices within earshot because apparently a device designed to perform possibly-sensitive actions on your behalf was assigned a model line wide, public audio trigger in order to make it feel more 'natural' or something; rather than some prosaic but functional solution like a trigger button/capacitive touch point/whatever; and that the device just silently fails stupid, no even informative feedback, in the even of server unresponsiveness or network issues. Both of these seem...less than totally fine...for something explicitly marketed for public use in crowded environments on what we euphemistically refer to as 'edge' network connectivity.
This. The "someone says 'Hey Siri/Okay Google' on TV/radio/loudspeaker" problem is a well-known failure mode, and if they don't have reasonable mitigation in place by now, they don't know what they're doing, and their product shouldn't be taken seriously. Whether that mitigation is blocking it during meetings, doing handshaking to limit commands to the nearest device when multiple nearby devices detect the hot word at exactly the same time, making it recognize your voice and not other random people's voic
Is this a glitch or a feature? (Score:2)
Hey Alexa... (Score:1)
Re: (Score:2)
If you're going to wear a powerful computer on your face, other people shouldn't be able to give it commands.
If more people DO give random commands to these stupid glasses, maybe the new glassholes will stop wearing them. Seems like a net win to me.
Not a Real World Issue (Score:5, Funny)
When the chef said, 'Hey, Meta, start Live AI,' it started every single Ray-Ban Meta's Live AI in the building.
Meta don't have anything to worry about. That can't possibly happen in the wild because it would require more than one person owning a set, and I've yet to meet anyone even considering purchasing them.
Re: (Score:2)
I do not like Zuck any more than the rest of you (Score:2)
but give the guy a break. This is hardly the first new product demo to have failed. Maybe they should have had the foresight to run it over a private network and not the center's WiFi, but knowing what will happen in the real world is very hard to get right.
Re: I do not like Zuck any more than the rest of y (Score:1)
Re: (Score:3)
Re: (Score:2)
In that case, the demo succeeded. I am absolutely sure one of the selling points at the C-Suite and a few of the more malicious advertiser and MBA crowd just beneath is that AI brings just enough pseudo-randomness while seeming useful to increase engagement and eyeballs.
This started getting noticeable with Google Search only a couple years after acquiring DoubleClick.
When it happens to Zuck, they fix it (Score:5, Insightful)
Not just a demo fluke (Score:2)
""When the chef said, 'Hey, Meta, start Live AI,' it started every single Ray-Ban Meta's Live AI in the building"
So anybody can control your Meta AI by speaking to it? Isn't that kinda a huge security issue?
Re: (Score:1)
""When the chef said, 'Hey, Meta, start Live AI,' it started every single Ray-Ban Meta's Live AI in the building"
So anybody can control your Meta AI by speaking to it? Isn't that kinda a huge security issue?
I know, right?
I'm just some random guy, and that's literally the first design scenario that pops into my head. For requirements, not a demo.
Re: (Score:2)
Re: (Score:2)
Re:Not just a demo fluke (Score:4, Insightful)
The problem isn't that they only had a small private server to handle all those requests. The problem is that the glasses were *all perfectly willing to execute.the verbal orders of a total stranger.*
Re: (Score:2)
So anybody can control your Meta AI by speaking to it? Isn't that kinda a huge security issue?
Obligatory xkcd [xkcd.com]
Wow (Score:2)
How ... could that not be considered?
I mean, I'm just some internet rando and it's like the first scenario that occurs to me (for freakin' requirements, not just for a demo).
Not Buying It (Score:5, Interesting)
I'm not buying their explanation (nor their product). Lag doesn't cause an AI to give you the wrong answer. The guy asked for the steps to make a meal and the AI responded as if half the steps had already been completed. He asked again and it gave the exact same response. A denial of service shouldn't cause an error like that unless for some stupid reason their software just repeats the last answer if it can't generate a new one. But if so then they also have major session management issues, giving some users the first steps and other users the next steps. For ease of use a dev server might not have proper user management setup, but they exposed it to the public... Expect a ton of hacking attempts at their next demo.
That example also points out multiple usability issues. The presenter interrupts the AI's responses multiple times implying that even they think the responses are too wordy. They are in control of the system. If responses are too fluffed up then stop doing that or add a more concise mode. It looks really bad when you're cutting off your own product. Another issue is the announcement triggering all the other devices. That's been a basic failure case from day 1 of voice input. They apparently haven't added anything to handle that. It's not too big a deal when the voice assistance is at a home, but smart glasses are meant to be used all the time. If all you have to do is yell a command they're going to have a major PR problem. It's less of an issue with phones as they're often locked and in your pocket, but a nearby glasshole is wide open. I guess that might technically be unauthorized computer use and land you in jail, but if you can make the claim that you were trying to delete all your own emails then you'd be off the hook. These systems don't seem to be voice locked to the user yet (nor to an activation button). My bank claimed identifying someone's unique voice print is easy and secure... (well to be fair I think AI killed those voice id checks a couple years ago, but maybe they still run them just without prompting you about it.)
The WhatsApp bug is more reasonable. (rant on) Linux can't do lockscreens properly either. Sometimes my Ubuntu Mate laptop locks when going to sleep and sometimes not. Often when unlocking, the desktop and its open windows are displayed for a moment prior to the login window blocking it out. From online reports, you can interact with the desktop during that shot timescale (which is plenty long for a newly plugged in, automatic input device to run commands). You can also do odd things like open the max amount of eom instances and after locking, the login window refuses to display so you can't log back in (switch to a different terminal, log in, then kill the screensaver to get your desktop back).
Re: (Score:2)
You're overthinking this. All he needed to do was appease shareholders. Many of those being investment bankers by profession... You know, those people that get the big bucks for investing no better than you if you flipped a coin.
His explanation points to something even they have heard of that ian't Facebook's fault. That is all that is necessary these days.
Re: (Score:2)
Re: (Score:2)
Any competent IT department would have planned for the large number of nodes connecting, and set up a separate network for the actual demo.
Hard to tell if it's incompetence, or lying, when you're dealing with incompetent liars.
Re: (Score:2)
Control-Alt-F (Score:2)
Will usually get you to the Linux console login from an X session screen locked or not (unless its been disabled in the X config). Obviously you need to know a login but even so, the lock is hardly bulletproof. No idea what wayland does, probably screen locking is another optional extra that needs to be implemented by the user or something.
Re: (Score:2)
Another issue is the announcement triggering all the other devices. That's been a basic failure case from day 1 of voice input. They apparently haven't added anything to handle that. It's not too big a deal when the voice assistance is at a home, but smart glasses are meant to be used all the time. If all you have to do is yell a command they're going to have a major PR problem. It's less of an issue with phones as they're often locked and in your pocket, but a nearby glasshole is wide open.
Can't wait to try this in public transportation: "Hey Meta, play "Never Gonna Give You Up"!"
Excuses.... (Score:2)
"funny reason their demo didn't work" says CTO (Score:2)
Subject Access Request (SAR) (Score:2)
Crappy tech is crappy tech ... (Score:2)
What else is new. If they cannot even get it right for the demo by the CEO, who in their right mind would buy this?
Popups/ads in your face? No thanks. (Score:5, Insightful)
Never mind the failed demo. I'm trying to imagine living with this technology if it DID work perfectly.
WhatsApp message notifications popping up constantly in front of your nose. Popup ADS you can't escape.
Sounds like hell to me.
Re: (Score:2)
WhatsApp message notifications popping up constantly in front of your nose. Popup ADS you can't escape.
Sounds like hell to me.
"We estimate we can sell up to 80% of an individual's visual field before inducing seizures." [ref [youtu.be]]
look at all the free press (Score:2)
yep, meta got a lot of free press over this, in addition to the normal articles. Everyone knows that demos do not always as planned. I have the same issue since whenever I want to show off my smart home, it almost always pick that time to act up. Just got to grin and bear it, then move on.
The real failure was no stand-alone backups (Score:1)
server load (Score:2)
"Meta's dev server wasn't set up to handle the flood of traffic from the other glasses in the building"
So a few glasses generate a "flood of traffic". It makes me wonder what kind of compute load these glasses impose on a server, and how much WiFi bandwidth they require in order to operate.
It's entirely possible that most of the functionality is implemented on the server side, and the glasses are just a dumb peripheral. All the video and audio from the glasses are sent upstream to Meta servers, which then t
Genius (Score:2)
Letting the side down (Score:3)
Do you wanna be part of the Evil Empire or not, 'cause any truly evil evil corporation would know how to fake a demo.
Obligatory XKCD (Score:2)
https://xkcd.com/1807/ [xkcd.com]
uh... wrong tree? (Score:2)
"When the chef said, 'Hey, Meta, start Live AI,' it started every single Ray-Ban Meta's Live AI in the building. And there were a lot of people in that building,"
The number of people isn't the problem here.
The "started every" is.
How did they not catch that during development and found a solution? I mean, the meme's where a TV ad starts Alexa and orders 10 large pizzas are a decade old now.