Related Items Go Here
Image credit: Meta
News

Meta Demoed Its New Smart Glasses AI Live On Stage, And It Went Terribly Wrong

Share

Meta’s presentation showcasing their brand-new augmented reality glasses went off track on Wednesday, as the live AI encountered glaring issues, prompting them to end the demonstration abruptly.

On Wednesday, Meta CEO and T-Pain collaborator Mark Zuckerberg showed off a range of new augmented reality glasses at the most recent Meta Connect 2025. In his presentation, Zuckerberg would claim these glasses to be the “first AI glasses with new resolution,” and would also boast a new hefty $799 version of its Meta Ray-Ban smart glasses.

However, the floor would fall out from underneath him (metaphorically speaking) as the glasses AI would encounter glaring issues in its live tech demo.

LiveAI demo fails on the first prompt at Meta Connect 2025. #Meta #AI #LiveAI

[image or embed]

— Shacknews (@shacknews.com) September 18, 2025 at 10:34 AM

In a painfully scripted setup, Mark would throw the stage to viral internet chef Jack Mancuso, who, wearing the new Meta smart glasses, would ask its live AI to help him make a Korean-inspired sauce for his steak sandwich.

“Hey, Meta, start Live AI,” Mancuso would start, moments before disaster. A long, awkward pause would then take place before the AI would eventually come online, cutting in, “I love this set-up you have here, with soy sauce and other ingredients.” Proving that it can, at least, scan objects.

Mancuso would ask if the AI could help him with his spur-of-the-moment Korean-inspired steak sauce for his sandwich, which the table had every ingredient laid out, ready to go (according to this online recipe at least).

The AI would list back the ingredients to Mancuso, saying he can make a steak-sauce sandwich with the ingredients he has laid out on the table. However, Mancuso would quickly interrupt the AI, asking what he should do first. The AI would then freeze for seconds, prompting Mancuso to repeat the question.

“You’ve already combined the base ingredients. So now grate a pear to add to the sauce.” Despite having all the ingredients ready to go and, of course, not having any of the base ingredients combined, now they suddenly needed to add a pear to the recipe.

The crowd started laughing in the background, prompting Mancuso to repeat his question for a third time. The AI would merely repeat that he should begin to grate a pear into the sauce since he had already combined the base ingredients.

Realising the tech demo was now wildly off course, Mancuso quickly blames the Wifi in the venue before throwing it back to Zuckerberg.

“The irony of the whole thing is that you spend years making technology and then the wifi of the day kinda catches you,” Zuckerberg would say.

Earlier this year, Meta invested $14 billion USD into Scale AI, a global workforce of contractors across Kenya, the Philippines and Venezuela who manually label images, text and video for machine learning applications. The justification for the investment was to secure high-quality training data for their own artificial intelligence models.

The result months later? Well, slapping big labels on soy sauce and sesame oil only to have Meta’s smart glasses acknowledge that the bottles were in fact, just that, raises some questions to say the least.

In July, Zuckerberg also announced that he will be spending billions in building massive AI data centres in the US. The first is expected to come online in 2026, and would cover an area nearly the size of Manhattan. Critics have noted that this would cause massive environmental and safety risks to the environment around the data centre.

The BBC would publish a story the same month, following Georgia resident Beverly Morris, and how her quality of living had significantly worsened since a data centre had been constructed nearly 400 yards away from her home.

 “I can’t live in my home with half of my home functioning and no water,” Morris would say. “I can’t drink the water”. She would show the BBC that even after fixing her water pressure and taps multiple times, the water still has residue inside it.

All of this to help fund AI glasses, which are emboldening perverts to take photos of people in public without their consent, while also introducing massive risks to individuals’ privacy and security. Can it at least tell you what you can cook after laying all the clearly labelled ingredients out for it to read? Apparently not just yet.

`