AI's Public Face-Plant: A Week of Unchecked Output

Illustration: chili pepper mascot illustrating public failures

AI's Public Face-Plant: A Week of Unchecked Output

This week, the digital world offered a masterclass in how not to deploy artificial intelligence. Three distinct, very public failures hit the internet, each a glaring example of companies shipping AI output without a single human bothering to check the obvious details. The public, naturally, caught it instantly. It wasn't subtle. It was a full-frontal collision with reality, proving that common sense still beats algorithmic ambition.

GEICO's Philly Follies

First up, GEICO decided to get "local" with some Philadelphia-themed billboards. The idea was simple: connect with the city's spirit. The execution, however, was a spectacular misfire, quickly flagged on Alex Holley's Instagram.

Imagine Independence Hall, one of America's most iconic buildings, standing proud but… missing its tower. Not a subtle omission, but a gaping architectural void where history should be. It’s like drawing the Eiffel Tower without its top, then calling it Parisian.

Then there was the Benjamin Franklin Parkway, a grand boulevard known for its museums and, crucially, Philadelphia City Hall. GEICO's AI decided to replace this architectural marvel with a generic church spire. A church spire. On the Parkway. It’s a geographical blunder so profound it borders on performance art.

And for good measure, errors cropped up in an Eagles locker room depiction. The details were off enough to trigger local fans, who know their team's sanctuary better than any algorithm. Commenters didn't mince words, calling it "lazy AI slop" and blaming "cost-cutting." The message was clear: GEICO skipped the human fact-check, and the AI delivered nonsense.

Just Eat's Culinary Catastrophes

Next, the food delivery giant Just Eat served up a bizarre menu of AI-generated images, brought to public attention by Dom Joly on Instagram. The goal was likely to make their online menus more visually appealing. The result was pure, unadulterated surrealism.

Take the "Jacket Potato" priced at £5.80. What did the AI conjure? Not a fluffy baked potato, but a literal jacket, covered in miniature baked potatoes. A sartorial starch statement. It’s a literal interpretation so absurd it makes you wonder if the AI has a secret life as a Dadaist artist.

Even more unsettling was the "Tiger Prawn" for £8.05. Instead of succulent crustaceans, the image displayed noodles topped with what appeared to be tiny, actual tiger cubs. Small, furry, endangered tiger cubs. On a plate of noodles. It was less a menu item and more a wildlife conservation nightmare.

The company's response was almost as baffling as the images themselves: "Every image had been manually checked." This statement, delivered with a straight face, directly contradicted the visual evidence. One can only assume the "manual check" involved a quick glance by someone who had never seen a potato, a prawn, or indeed, a tiger. Or perhaps they were just checking for offensive content, missing the obvious, hilarious, and deeply weird.

Unitree's Robotic Rumpus

Finally, we turn from digital blunders to physical failure, courtesy of Unitree's H2-X humanoid robot. Captured in two Facebook reels, this incident offered a stark reminder that even the most advanced hardware can stumble.

The robot, designed for impressive feats of agility, attempted a martial-arts aerial flip. A complex maneuver, requiring precision and balance. It launched, it spun, and then it landed. Or rather, it failed to land. With a distinctly ungraceful thud, the robot crumpled.

The most concrete detail? Its lower leg detached. Not a minor wobble, not a software glitch, but a literal piece of the robot lying on the floor, separate from the main body. The future, apparently, has detachable limbs. It was a raw, unedited moment of mechanical vulnerability, instantly shared and dissected by an online audience. No company spin could hide the stark reality of a robot literally falling apart mid-demonstration.

The Discourse

The immediate aftermath of these failures saw a predictable pattern of voices.

On one side, the public. Armed with smartphones and a healthy dose of skepticism, they became instant quality control. Their voices, often sharp and humorous, echoed across social media. They called out "lazy AI," "cost-cutting," and the sheer absurdity of the errors. They didn't need a technical manual to spot a missing tower or a tiger cub on their dinner plate. Their collective eye for detail proved far superior to whatever internal checks these companies had in place. "Did anyone even look at this?" was the unspoken question hanging over every comment section.

On the other side, the companies. GEICO remained largely silent on the specific billboard blunders, letting the internet have its fun. Just Eat, however, offered the baffling claim that "every image had been manually checked." This response, rather than reassuring, only fueled further disbelief. It suggested either a profound misunderstanding of what "checking" entails or a deliberate attempt to downplay the lack of oversight. Unitree's robot failure, being a physical event, spoke for itself, leaving little room for corporate spin beyond acknowledging the inherent challenges of robotics.

The core of the discourse centered on trust and competence. When companies deploy AI without human gatekeepers, they risk not just embarrassment, but a significant erosion of public confidence. The public expects a baseline level of quality and common sense, and when AI fails to deliver, the blame falls squarely on the human decision-makers who allowed it to go live.

The Lesson Learned (Hopefully)

These aren't isolated incidents. They are symptoms of a broader trend: companies pushing AI-generated content and functionality into the public sphere without sufficient human oversight. The allure of automation, the promise of efficiency, often overshadows the critical need for a human eye, a local expert, or simply someone with common sense to review the output.

The public isn't fooled. They possess local knowledge, cultural context, and an innate ability to spot the ridiculous. They are the ultimate quality control, and they are quick to call out "lazy AI" when they see it. The cost of "saving money" on human checks isn't just a few viral laughs; it's public embarrassment, brand damage, and a dent in consumer trust.

AI is a powerful tool, but like any tool, it requires skilled operators and rigorous quality assurance. These public failures serve as a stark reminder that while algorithms can generate, they cannot yet discern context, cultural nuance, or the simple fact that a jacket potato is not a jacket. Until that changes, a human in the loop isn't just good practice; it's essential for avoiding public humiliation.

Sources and further reading

GEICO Philly-style billboards, via Alex Holley Instagram Just Eat, via Dom Joly Instagram Unitree H2-X humanoid robots, via Facebook Unitree H2-X humanoid robots, via Facebook