DEV Community

Cover image for Humyn Labs Bags $47M to Teach Robots Using Human Movements
Masood Akhtar Vaheed
Masood Akhtar Vaheed

Posted on

Humyn Labs Bags $47M to Teach Robots Using Human Movements

Humyn Labs' Big Bet: Real People, Real Data, Real Money

$47 million raised, $50 million ARR in half a year, and founded just months ago: these are not your run-of-the-mill startup numbers. Humyn Labs has arrived on the scene like a bat out of hell, and with backing from heavyweight investors like Accel, Lighthouse Canton, and Trifecta Capital, they are not exactly sneaking in quietly. You do not see too many companies these days claiming they will teach robots by straight-up copying human actions on camera, then pulling in revenue that fast. Something is either working very right, or there is a touch of smoke and mirrors somewhere.

Why "Physical AI" Suddenly Matters

So, what is all the fuss about? Physical AI, as Humyn Labs defines it, is basically about teaching machines by having them watch and learn from humans. It is less about GPTs spitting out text and more about robots figuring out how to mop a floor or pick apples after seeing how people do it. Humyn Labs says they have set up shop in over 20 countries, aiming for data that is not stuck in a lab somewhere in Silicon Valley but comes from people actually living their lives, whether that is in a kitchen in Brazil, a factory in South Korea, or a street in India.

The company has built a hardware system with two cameras to record folks doing everyday stuff. That video is then processed and labeled by their own software. The end result is a giant, tagged dataset that helps robots get better at real-world tasks. The more diverse the data, the less likely you are to end up with a robot that can only clean American toilets or sort apples the European way. If you want a robot that understands the chaos of a Mumbai street or the speed of a Korean assembly line, you need to collect data where the action is.

The Catch: Privacy and the Human Factor

Of course, recording people all day is not exactly a feel-good, no-strings-attached move. There is a real tension here: on one hand, you need genuine, messy human movement to build useful AI; on the other, you have privacy, consent, and all the messiness of data ownership. It gets a bit awkward when you realize some of these cameras are running in homes and factories, spaces people generally like to think of as private.

There is also the elephant in the room: workers, sometimes in lower-income settings, are being asked to record themselves doing the very tasks robots might eventually take over. It is like if textile workers were asked to design the machines that could automate their own jobs. There is no easy answer here, but pretending it is not a dilemma feels a bit too convenient.

The Global Play

With operations in places like San Francisco, London, and Seoul, Humyn Labs is clearly thinking big. Expanding into Brazil, South Korea, and India is not just about cheap labor or new markets. It is about training robots using data from as many environments as possible, because a robot that can only handle an American-style kitchen is pretty limited. This sort of global scale comes with headaches, of course. Setting up the hardware, managing compliance in all those regions, handling languages and legalities, it is far from trivial.

What stands out is the brute-force approach: capture as much raw, unsanitized human behavior as possible, wherever you can. It is bold, maybe a bit aggressive, but it is also the kind of thing that can set a company ahead of slower, more cautious competitors.

What Comes Next?

Humyn Labs is moving fast, chasing growth and attention from investors. But beneath the big numbers and global footprint, there is a question of where the line gets drawn between innovation and intrusion. If recording the world is how these companies build the future of robots, who really gets to decide what is fair? Maybe the real test will not be technical at all, but whether people are willing to let the physical world get recorded, one awkward camera setup at a time.

Source: https://inc42.com/buzz/humyn-labs-47-mn-round-from-blinkit-to-vice-city-more/

Top comments (0)