They may seem like scenes out of a slapstick comedy, but viral robot mishaps are no laughing matter — experts even warn they could foreshadow a “Terminator”-level apocalypse.
A humanoid dance bot was recently performing for patrons at the Haidilao hotpot restaurant in San Jose, California, only to end up literally tearing up the dance floor, knocking over tableware, smashing plates, and sending chopsticks flying like Mr. Magoo’s machine doppelgänger.
The spectacle ended with human staffers dragging the flailing droid out the door as bemused customers looked on.
While the malfunction elicited guffaws online, techsperts warn that these cybernetic pratfalls could signal something more alarming as bots become increasingly embedded in everyday life, as these consumer-grade machines can come apart when the rubber hits the road.
“I think these incidents are often treated as funny only because the immediate harm was limited and the context was theatrical,” Dr. Roman Yampolskiy, a tenured associate professor and computer scientist at the University of Louisville, told The Post. “People laugh at low-stakes failure.”
But, the AI specialist added, “From a safety perspective, they should also be taken seriously, because they reveal something important: systems that appear polished and entertaining can still behave unpredictably in the physical world.”
In the past couple of months alone, a handler in China was kicked in the groin by an advanced Unitree robot he was controlling, and a droid shockingly slapped a child during a dance demo gone awry.
What if, Yampolskiy inquired, similar malfunctions occurred around a baby, a hospital patient, or a member of the public during a police interaction?”
“The event would be viewed not as comic relief but as a dangerous systems failure,” said the researcher, who has authored over 100 papers on the existential threat of AI. “A glitch in a dancing robot is mostly embarrassing. A glitch in a security robot, delivery system, self-driving platform, medical assistant or industrial machine can injure people, damage property or trigger cascading failures.”
While seemingly far-fetched, Yampolskiy said that these minor hiccups scale into larger problems as the automation offensive makes AI more ubiquitous in sectors like security, healthcare and even romance.
More than 60 bomb squads across the US and Canada are already using Spot, Boston Dynamics’ hyper-advanced, 75-pound robo-dog for roles ranging from armed standoffs to hostage rescues, Bloomberg reported.
Multiple firms are also working on developing hyperrealistic helper bots for in-home use, such as Clone Robotics’ “Protoclone” — touted as the “world’s first bipedal, musculoskeletal android” — which can allegedly walk, talk and complete chores.
“As AI moves from screens into bodies and institutions, the cost of error rises dramatically,” Yampolskiy declared.
The cutting-edge automatons are becoming stronger and faster, too.
Researchers in South Korea have developed a chemical structure for an artificial muscle that can potentially allow humanoid robots to lift 4,000 times their weight, while China’s Bolt humanoid robot can run up to 22 m.p.h.
The bots can kick our butt as well.
Viewers were understandably concerned over a video of Unitree’s next-gen humanoid robot, the H2 — touted as having both commercial and personal use — that was spotted lifting a smaller droid off the ground with a knee strike, sending its breastplate flying.
What would happen if a human were on the receiving end of such a hit? Such mishaps aren’t just theoretical. In February, a Unitree G1 robot accidentally struck a man in the nose, causing him to bleed, while it was trying to right itself after falling during a performance in China.
“With existing reinforcement learning policies, their robot is trained to do whatever it takes to stand up after a fall,” Eren Chen, who claims to work for the robotics firm Booster Robotics, wrote on X. “During that recovery attempt, it kicked someone in the nose, causing heavy bleeding and a possible fracture.
“This should be treated as a high-priority safety issue for Unitree to fix.”
However, Yampolskiy pointed out, “human safeguards can reduce risk, but they do not eliminate it.”
“Better testing, physical constraints, geofencing, kill switches, supervision, and strict deployment standards all help,” he declared. “But no complex system is perfectly reliable, especially when it operates in open-ended real environments.”
It comes with the technological territory.
“So, yes, some accidents are to be expected, just as with cars or other powerful tools,” he said. “The difference is that society must decide what level of failure is acceptable, and that threshold should be very low for systems operating around people.”
Unfortunately, tech bigwigs haven’t always been forthcoming about potential pitfalls — or at least haven’t used them as signs they should slow their roll.
Robert Gruendel, a former engineer for human robotics firm Figure AI, sued the company because, he claimed, they fired him for warning that their robots “were powerful enough to fracture a human skull,” CNBC reported.
However, Figure has denied the allegations, with reps saying that Gruendel was “terminated for poor performance,” and that his “allegations are false.”
Who is culpable should one of these cybernetic Frankensteins go haywire?
Yampolskiy believes that the “primary responsibility lies with the companies that design, deploy and profit from these systems.”
“Developers, operators and users may each bear some share, depending on the facts, but the burden should fall most heavily on those who choose to release insufficiently reliable systems into real-world settings,” he said. “More broadly, these episodes are early warning signs.
“Today, they go viral as odd or amusing clips. Tomorrow, with more capable and more widely deployed systems, the same class of failure may be discussed in terms of injury, liability, and public safety.”
