A new wave of humanoid robots is grabbing headlines, but experts warn the promises are ahead of real-world performance.
Humanoid robots are popping up everywhere lately, from filmed stunt “appearances” abroad to high-profile demonstrations at U.S.. events.. Yet behind the spectacle. a central question is resurfacing in the tech world: are humanoid robots mainly hype. or are they genuinely moving toward everyday reality?. Misryoum
In a discussion featured on the “Today. Explained” podcast. tech writer and journalist James Vincent—who previously wrote a Harper’s Magazine cover story titled “Kicking Robots”—described firsthand meetings with humanoid robots and how quickly the emotional punch of the demonstrations collides with the practical limits of the technology.. Vincent said he met robots from two U.S.-based companies. including Apptronik and Agility Robotics. each building distinct styles of humanoid designs even though both resemble humans with arms and legs.
According to Vincent. Agility’s robots are aimed primarily at warehouse environments and are built with a look he described as more inhuman. while Apptronik makes a robot that is designed to appear closer to human proportions and stands upright so that it can be seen more directly “eye to eye.” He said he shook hands with the machines and even played a version of rock-paper-scissors with a robot. but the moment that stuck with him involved pushing one—an encounter that captured the uncanny closeness humanoid robots can achieve.
Vincent said he wasn’t allowed to physically kick a prototype robot for safety reasons. but he was instead given a stick-like tool with safety foam taped on the end and asked to push it.. He described how the robot wobbled, then corrected its balance with a human-like recovery, and then returned to face him.. To Vincent. that behavior was unsettling precisely because it looked so natural. reinforcing the sense that these machines are no longer just futuristic concepts.
The conversation then shifted to what humanoid robots are actually being sold to do.. Vincent said the central pitch is that humanoid robots are meant to perform “anything that an able-bodied human can do. ” including routine workplace tasks such as sorting packages. bolting parts in industrial settings. and other manual labor.. But he also emphasized the challenge in that promise. describing the expectations as a “big ask” when measured against the complexities of everyday physical work.
Vincent pointed out that demand for humanoid robots is coming largely from companies in the U.S.. and China, with big tech firms increasingly entering a space that earlier relied more heavily on startups.. He noted that Meta has bought a robotic startup and that Google has long worked on robots while testing aspects of its AI on robotic systems.. He also discussed Tesla’s involvement. framing it as part of Elon Musk’s broader push—where Tesla’s humanoid robot project. Optimus. is presented as a potentially highly profitable product.
A key driver behind the sudden acceleration in humanoid ambitions. Vincent said. is AI—specifically the boom associated with ChatGPT and the rise of deep learning systems.. He explained that large language models and chatbots have made it easier to imagine transferable technology for physical machines. suggesting that systems trained on large amounts of data could learn how to perform tasks by recognizing patterns rather than relying entirely on manual programming.
In the past. Vincent said. robots often required detailed instructions at a mechanical level—commands specifying how arms should move. how far they should travel. and how much force they should apply.. The newer approach, as he described it, involves training models that learn connections between inputs and desired outputs from data.. Companies. he said. are betting that with enough training and data. they can “solve the problem of physical robotics. ” producing machines with the dexterity to handle many different tasks.
Still. the criticism Vincent raised is rooted in the difference between talking to a chatbot and controlling a machine in the physical world.. While chatbots can sometimes make mistakes, those errors remain largely in text.. In contrast. when physical robots make errors. the consequences can become more dangerous. a point that complicates claims that the same learning breakthroughs can be directly ported into real-world mobility and manipulation.
Vincent also addressed the growing emphasis on placing humanoid robots in the home. where companies market them as an all-purpose helper—described as a “perfect robot butler” that could handle chores such as dishes and laundry.. He argued that the tolerance for error is fundamentally different in domestic use. where mistakes that might be minor in a conversation—like a chatbot misunderstanding a request—could become unacceptable if a robot breaks objects or fails repeatedly.
In comparing China’s development pace with the U.S.. Vincent said the main difference is speed and scale. with China moving quickly and. in his view. more effectively.. He tied part of that momentum to demographic pressure. noting China’s rapidly aging population and how planners there may see humanoid robotics as a way to address both gaps in manufacturing labor and rising demands in social care.
Vincent said Chinese state planning could view humanoid robotics as fitting into those needs, while the U.S.. approach has placed more emphasis on home products as a marketing tool targeted at wealthier consumers.. He also pointed to a core economic advantage China has that the U.S.. does not to the same degree—scale in manufacturing—suggesting the ability to produce thousands of units at a time is a reason China can pull ahead.
Ultimately. Vincent said he sees the trajectory as closer to the progress associated with “flying cars” than the fully realized vision of a near-instant transformation comparable to chatbot performance.. He acknowledged that capabilities have advanced quickly and that there has been a real leap forward. but he argued that this does not match the high expectations pushed by prominent executives and companies promising rapid deployment at home.
He cautioned that some public claims go beyond what the technology can reliably deliver. including promises of year-ahead home use that guarantee the robot will never make mistakes and won’t cause harm.. In his view. while humanoid robots may become more common over the longer term—both in work settings and within homes—the timeline for widespread. reliable use is likely to be measured in decades rather than the short windows suggested by the most aggressive marketing.
Vincent said that in the next few years. he doubts the hype will be realized. even as he expects humanoid robots to grow more present over the next 10-plus years.. For now. his central theme is that the demonstrations are real. and the progress is tangible. but the gap between what humanoid robots can do today and what they are being promised to do tomorrow remains wide enough to reshape expectations—especially when physical safety. domestic quality. and everyday reliability are on the line.
humanoid robots AI robotics Optimus warehouse robots home robotics China robotics Apptronik Agility