The implication from new demonstrations of humanoid robots dismantling utensils or assembling cars is that replicating human limbs with single-purpose robot arms is the old way of doing automation. The new approach aims to replicate the way humans adapt when they think, learn and work. The problem is that a lack of transparency about the human labor involved in training and operating such robots leads the public to misunderstand what robots can actually do and fail to notice the strange new forms of work around them.
Consider how, in the AI age, robots often learn from humans who demonstrate how to do a job. Making use of this data at scale is now moving forward black Mirror-esque scenario. For example, a worker in Shanghai recently spent a week wearing a virtual-reality headset and an exoskeleton while opening and closing a microwave door hundreds of times a day to train the robot next to him, rest of the world informed. In North America, robotics company Figure is planning something similar: It announced In September it will partner with investment firm Brookfield, which manages 100,000 residential units, to capture “huge amounts” of real-world data “across a variety of home environments”. (Chitra did not respond to questions about this effort.)
Just as our words became training data for large language models, our movements are now set to follow the same path. Except that humans may face an even worse situation in the future, and that has already begun. Roboticist Aaron Prather told me about recent work with a delivery company whose employees wore movement-tracking sensors while moving boxes; The data collected will be used to train robots. The effort to create humanoids will require manual laborers to act as data collectors on a large scale. “It’s going to be weird,” says Prather. “It doesn’t have any message.”
Or consider tele-operation. Although the endgame in robotics is a machine that can complete a task on its own, robotics companies employ people to operate their robots remotely. NewStartup 1X, a $20,000 humanoid robot, is set to be shipped into homes this year, but the company’s founder, Bernt Øyvind Bornich, recently told me he’s not committed to any set level of autonomy. If a robot gets stuck, or a customer wants it to perform a difficult task, a tele-operator from the company’s headquarters in Palo Alto, California, will operate it, watching through its camera for things like ironing clothes or unloading the dishwasher.
This isn’t inherently harmful – 1X gets customer consent before switching to tele-operation mode – but privacy as we know it won’t exist in a world where tele-operators are operating in your home via robots. And if home humanoids are not truly autonomous, the arrangement is better understood as a form of wage arbitrage that recreates the dynamics of gig work while, for the first time, allowing physical work to be done wherever labor is cheapest.
We have walked similar roads before. Performing “AI-powered” content moderation on social media platforms or gathering training data for AI companies often requires workers in low-wage countries to view disturbing content. And despite claims that AI will soon take sufficient training on its output and learn on its own, even the best models require a lot of human feedback to work as intended.
These human workforces don’t mean that AI is just vaporware. But while they remain invisible, the public continues to overestimate the true capabilities of machines.
It’s great for investors and publicity, but it has consequences for everyone. For example, when Tesla marketed its driver-assistance software as “Autopilot”, it raised public expectations about what the system could do safely – perverting the Miami jury. recently found Contributed to an accident that killed a 22-year-old woman (Tesla was ordered to pay $240 million in damages).
The same would be true for humanoid robots. If Huang is right, and physical AI is coming to our workplaces, homes, and public spaces, then the way we describe and examine such technology will matter. Yet robotics companies are as opaque about training and tele-operation as AI companies are about their training data. If this doesn’t change, we risk mistaking hidden human labor for machine intelligence – and see far more autonomy than actually exists.
