Every few weeks, it seems, a certain type of video will make the rounds on the internet. Promoted by one of a number of robotics firms, it features a humanoid—a robot designed to resemble a human—performing impressive feats otherwise thought to be unique to our species. The robot might be watering plants, say, or loading a dishwasher, with an uncanny precision that inspires equal parts wonder and terror.

The buzz around humanoids, writes James Vincent in our December cover story, has become unavoidable: “There are headlines about Chinese bots running half marathons, ominous videos of muscled humanoids twitching on gantries, clips of robot fight clubs.” But to what degree do these publicity campaigns reflect the current state of humanoid technology? Vincent visited several robotics firms in an attempt to distinguish between hype and genuine technological innovation. Does the heavily capitalized humanoid industry represent little more than a marketing-fueled bubble? Or does it point to an inevitable future?

I spoke to Vincent about the opaqueness of the robotics industry, the deceptivenessof its marketing, and just how frightened we should be of a humanoid future.

Matthew Sherrill: Reporting on humanoids involves a number of relatively unique challenges. You mention in the article that in the decade you’ve spent reporting on the tech industry, executives have grown increasingly cagey, or even antagonistic, toward journalists. This seems particularly true of the robotics industry. Can you talk about the difficulties you encountered persuading robotics firms to speak with you—let alone grant you access to their robots?

James Vincent: Practically speaking, these companies just don’t have a lot of robots on hand, so freeing one up to show it to a journalist can genuinely mess with their schedules. Tactically, they’re careful about what they show, and to whom. Sometimes, this is because they don’t want information to leak to rivals, but more often it’s about controlling the narrative, and avoiding bad stories that can spook investors and public confidence.

The company I had most trouble speaking to was 1X, which has been incredibly ambitious in its promotion of a home robot butler that they claim will be fully operational in a few years. For reasons that aren’t clear to me, 1X decided to give me the runaround, and eventually ignored me entirely. They’ve gone on something of a media blitz since this story went to print and the company began accepting preorders for their humanoid, NEO, but at the time they were very limited with media engagement, choosing to show their robots only to individuals and outlets which gave them what I consider a biased appraisal, emphasizing the tech’s long-term potential and downplaying its limitations. I don’t think this is necessarily a terrible thing to do, but it suggests a lack of confidence. If your robot can do everything you say it can, why cherry-pick coverage?

MS: Many of the tech demonstrations that robotics firms are willing to put on for journalists are highly scripted affairs that try their best to look unrehearsed or spontaneous. You quote one roboticist who compares the performance of many humanoids to a “blind gymnast,” unaware of the actual environment in which it’s operating. How difficult is it to determine how impressive a particular piece of robotics technology is, when the big firms are only willing to show off their products under tightly managed conditions?

JV: It’s very hard to judge the credibility of this tech, and your ability to do so depends on the circumstances that you’re seeing it operate under. So, if you’re getting an in-person demo, you can usually trust what you see, because it’s obvious when a robot fails to adapt to new environments or objects, and you can ask the company directly about the tech’s limitations. The reputational penalty for outright lying to a journalist—especially about something like functionality—is still significant enough that companies are generally going to tell you the truth.

But, in scenarios like prerecorded demos or press events, firms can really obscure their products’ true capabilities. Cutting-edge humanoids, like a lot of tech powered by AI, are “jagged,” meaning they excel at specific tasks but have trouble generalizing the underlying functionality. So, you get robots that can backflip like Simone Biles but can’t flip an egg in a frying pan. Companies use this to their advantage. Not by lying, exactly, but by encouraging audiences to see what they want them to see: superhuman machines.

MS: In the piece, you touch on the role China is playing in the humanoid boom. In most respects, it seems like the country is miles ahead of the rest of the world, and yet, for a host of reasons, it’s even more difficult to discern just how advanced its humanoids are. At the same time, they’ve staged a number of humanoid-focused spectacles, like the World Humanoid Robot Games, that, to my amateur eye at least, look pretty impressive. What role do you see China playing in our humanoid future?

JV: Unless there are major changes in the United States, it’s almost certain that China will lead the development of this technology in the twenty-first century. China has an incredible ecosystem of manufacturers; it has generous state funding and large-scale production; and, most importantly, it has the will and energy to push this tech forward—helped by these big events like the Robot Games. But, in terms of actual capabilities, there’s not a huge gap between the U.S. and China right now. In fact, the U.S. still has the edge when it comes to AI development and autonomous systems, where firms like Google and Tesla are the world leaders. But because scale is so important for developing this technology, China is on course to overtake the U.S. soon. However, everything I’ve said about the jagged capabilities of robots and selective demos applies to Chinese bots, too. Seeing a robot sprint past a human is impressive, but it doesn’t mean that the robot can do anything but run very fast. There’s a temptation to credit China with breakthroughs simply because of narratives about the “Chinese century” and so on, but skepticism should not stop at home.

MS: While we were working together on this piece, I was interested in your resistance to a certain type of blanket skepticism toward the humanoid industry. You witnessed plenty of technology that you found genuinely astonishing. Can you share some of the moments that most surprised or impressed you?

JV: Skepticism is vital, but I’m sometimes bothered by what I see as knee-jerk criticism. You get it a lot in discussions of AI, where people are culturally and politically opposed to the companies pushing this tech, with good reason, and so they overlook genuine achievement. We don’t tend to see our own biases, myself included, and I’m sure some people will think I’m credulous. But in order to be effective, any critique has to be realist, which means acknowledging when advances are made. I think this is particularly important in progressive circles, which have ceded authority in discussions on these matters because of a failure to accurately account for what these systems can and can’t do.

I’d say there is something about seeing these machines in the flesh, as it were, that is awe-inspiring, frankly. The most impressive moment for me was when I tried to knock down a humanoid built by Apptronik. There was something about being able to feel the robot’s size and weight in response to my actions. I was standing eye-to-eye with it, and it reacted to my shoves just like a human would—managing to stay perfectly upright by moving its arms and legs to counter changes in momentum. Oddly, it reminded me of the first time I saw a gorilla in a zoo. I obviously knew what a gorilla looked like before this and understood, in theory, that this is an animal that shares an evolutionary past with humans. But when you see one in person, even just sitting about or eating, your brain starts making all these subconscious assessments, like, “Oh, it’s strong. Oh, it’s smart.” And then, “Oh, it’s thinking.” With robots we can’t extrapolate capabilities from casual observation in the same way we can with other species, but there’s no denying that these machines are real things capable of effecting real change in the world.

MS: Assuming the widespread adoption of humanoids on some time scale, what most frightens you about that future? Unemployment? Mass uprising? Something more banal?

JV: I’m definitely not worried about a robot uprising in which machines become an intelligent, independent faction in society, but I do think there could be an uprising in a different sense. What worries me is how the automation of manual labor could massively change social dynamics. To be as clear as possible: humans create value through labor, and this gives the working classes leverage against the owners of capital. If this value is taken away, how does the working class retain its agency?

Now, it’s fair to say that automation has been happening for centuries. But what’s scary about this current moment is the potential for humanoids to automate a huge range of work incredibly rapidly—in a matter of years, perhaps, if the tech really takes off. The bet that many of these humanoid firms are making is that their development will follow a similar path to that of AI. There, in only a few short years we went from AI systems that could barely string a sentence together to models that can produce text that is, for many, indistinguishable from human writing. We’re already seeing how this tech is affecting white-collar industries, with mass layoffs at tech companies, and I think the disruption caused by the automation of blue-collar work will be much more destructive. In part this is because of the nature of the work, but also because of our political environment. Governments like the United States are unwilling or unable to constrain the power of the capital classes, and humanoid automation will only strengthen their hand. The future that scares me is not a robot uprising led by robots, but one led by billionaires. And that’s the direction we’re currently heading in.

MS: Let’s imagine that we can somehow circumvent this kind of dystopian future. What does a best-case humanoid scenario look like?

JV: There’s real potential for humanoids to relieve us of drudgery, as other technologies have in the past. I’m no technophobe, and I don’t think it’s naïve to celebrate the fact that certain forms of work have become less common in recent centuries. I’m talking about backbreaking labor—the sort that tears up human bodies, and with it, human potential, to break rocks, harvest wheat, and so on. If humanoids become cheap enough and capable enough, they could handle more of this work, and that’s a good thing. The health-care sector has good potential here, for example. If robots can move supplies around a hospital or change sheets, then you can free up people to do work that actually requires human interaction and empathy. All of this depends on broader social dynamics, on respecting these jobs and making sure people aren’t simply squeezed into lower wages, but I think that balance is possible to achieve.

MS: Possible—but is it likely?

JV: In ten years’ time? I doubt it. But in fifty? All bets are off.