Breaking the humanoid robot delusion
A Silicon Valley company called 1X this week announced a humanoid robot that does all your housework.
The robot is called NEO. The company says NEO is the world’s first consumer-ready humanoid robot for the home. It is designed to automate routine chores and offer personal help so you can spend time on other things.
The company’s odd, 70’s style demonstration video shows NEO doing a long list of tasks, such as folding laundry, organizing shelves, taking out the trash, and cleaning up rooms. It also performs basic functions such as opening doors, fetching items, and turning off lights, and is shown dancing and hanging out — acting as a kind of friend.
NEO operates by navigating homes autonomously and learning new skills through updates and continued use. It can be set up with just a button press or a simple voice command, according to the company. If the robot can’t do something, the owner can schedule a session with a 1X expert, who will remotely guide the robot through the task, letting NEO learn and get the job done at the same time.
NEO’s AI enables users to interact with it in natural conversation without relying on screens. It can listen and respond only when spoken to directly. Visual intelligence lets NEO recognize objects and context, like reading ingredients laid out on a counter and suggesting recipes. It remembers past conversations and user preferences so it can adapt over time, manage reminders, help with scheduling, keep grocery lists, and track the user’s learning progress.
The robot’s AI is trained on real-world data to allow it to adapt to the variety of home environments it will encounter.
Its hardware is based on a tendon-driven design using high torque motors for smooth, safe movements, allowing it to operate safely among people. NEO features hands with 22 degrees of freedom, a soft body made from 3D lattice polymer, and weighs 66 pounds. It can lift up to 154 pounds and carry up to 55 pounds, while producing just 22 decibels of noise, which is less than a modern refrigerator. NEO comes equipped with Wi-Fi, Bluetooth, 5G connectivity, and speakers built into its pelvis and chest, designed so it can even be used as a mobile entertainment system.
NEO’s design is meant to blend in with typical home interiors, with neutral colors and a fabric “suit” and shoes. It comes in tan, gray, or dark brown and is now available for pre-order through 1X’s online store. The first units are slated to ship for US-based customers in 2026, with 1X planning to expand to other markets in 2027. The reserved price: $20,000 for early access with priority delivery next year. (There’s also a $499-per-month subscription model.)
The company’s CEO and founder, Bernt Børnich, said in a press release that NEO marks the moment when humanoid robots, once only seen in science fiction and research, become real consumer products that any person can own.
Wow! So, the era of home humanoid housekeeper robots is finally here! The Rosey the Robot dream from the Jetsons has finally been realized.
Except it hasn’t.
The trouble with autonomous humanoid robots
The announcement and slick video showing NEO is smoke and mirrors. The robot can autonomously perform almost none of the feats shown in the video. (In fairness, the company doesn’t claim that it does, but wildly downplays the amount of improvement needed in the future to enable autonomous action.)
The video was created almost entirely through teleoperation. That means somebody wearing a VR headset and special equipment was actually folding laundry, putting away the dishes and carrying trash, and the robot was acting as a puppet, aping the movements of the person remotely controlling it.
If that scenario sounds familiar, then you may recall the “We, Robot” event held by Tesla a year ago to showcase cars at Warner Bros. Discovery’s studio lot. At the event, Tesla Optimus robots mingled, poured drinks, danced, and even sang “Happy Birthday” to attendees. The robots seemed to take orders, answer questions, and hand out beverages capably.
In reality, the robots were teleoperated by Tesla employees who were stationed out of sight. When people were impressed by a robot’s ability to converse, they didn’t know they were chatting with some employee.
Some technology CEOs, including Elon Musk and Bernt Børnich, practice what you might call “faith-based innovation.” They hype AI and robot capabilities that do not exist.
In the case of the NEO, 1X is already accepting money based on the promise that researchers will invent capabilities very quickly.
And in the case of Tesla, they’re not yet accepting money for Optimus robots. But they are definitely accepting money for cars that the CEO promises will get self-driving capabilities very soon.
At the Nvidia GPU conference in San Jose on March 17, 2015, Musk said full autonomy was a “solved problem” and would arrive by 2018. In late 2016, he said Teslas could drive coast-to-coast without human input by the end of 2017 and that every Tesla made from then on would be capable of “level 5 autonomy” within two years. During 2018 and 2019 earnings calls and events, he promised “full self-driving features” would be enabled via software in “three to six months” — and said a million Tesla robotaxis would be operating autonomously by mid-2020.
At Tesla’s Autonomy Day on April 22, 2019, Musk claimed that by the end of 2020, Teslas would be able to drive themselves safely enough that users “could fall asleep” and that regulatory approval would follow within a few years. And in interviews throughout 2020 and 2023, including tweets and earnings calls, Musk continued to say that full autonomy was “very close” and he expected completion within the year.
All these promises were wrong.
In fact, the technology for full self-driving using only cameras and AI has not been invented yet, and it’s completely unknown how many months, years or decades that invention could take. (Waymo cars, as well as most other self-driving cars, use cameras, lidar, and radar together to build a more detailed and redundant 3D map of the environment.)
As self-driving abilities, the humanoid robot game is rife with smoke-and-mirror demonstrations that create the illusion of autonomy, but in fact are mainly remote controlled:
Figure AI’s Figure 03: Videos of Figure 03 performing at-home tasks went viral for their realism and fluidity, but the company has been scrutinized for whether these were fully autonomous or tele-operated, as similar robots (like NEO from 1X Technologies) often use significant tele-op components, especially in publicity demos.
Unitree G1: Unitree’s humanoid robot has been demonstrated running and performing dynamic movements. At events like CES, Unitree staff allowed attendees to remote-control the G1, suggesting that some public demos are human-operated.
Reachy by Pollen Robotics: Reachy can operate in a very limited way autonomously or be tele-operated with VR headsets for precise remote control, a feature highlighted in several high-profile demos and videos where observers may have believed the robot was acting independently.
Reflex Robotics, Watney Robotics, and others: These companies employ teleoperation for tasks like cleaning, organizing, or even novelty “robot gaming,” where the humanoid form factors are presented as capable, but rely on remote human input for the most sophisticated actions.
Why robot companies blur the line
Full autonomy in perceiving, planning, and manipulating like a human is a massive technology challenge. Robots have to be meticulously and painstakingly trained on every single movement, learn to recognize every object, and “understand” — for lack of a better word — how things move, how easily they break, what goes where, and what constitute appropriate actions.
One major way humanoid robots are trained is with teleoperation. A person wearing special equipment remotely controls prototype robots, training them for many hours on how to, say, fold a shirt. Many hours more are required to train the robot how to fold a smaller child’s shirt. Every variable, from the height of the folding table to the flexibility of the fabrics has to be trained separately.
While these robots are being trained, it’s all very impressive if you can’t see the person controlling the robot.
The temptation to use impressive videos of remotely controlled robots where you can’t see the person controlling them to raise investment money, inspire stock purchases and outright sell robot products, appears to be too strong to resist. Realistically, the technology for a home robot that operates autonomously the way the NEO appears to do in the videos in arbitrary homes under real-world conditions is many years in the future, possibly decades.
But that doesn’t stop some companies from selling you products now on capabilities that haven’t yet been invented. Breaking the humanoid robot delusion – ComputerworldRead More