Why humanoid robots and not something more

Someone asked me this at a dinner party last week. “Why are they building robots that look like people? Wouldn’t wheels be better? Or four arms?”

Yes. Wheels would be more stable. Four arms would be more efficient. A snake-shaped robot could fit into spaces a humanoid never could. A flying drone can reach places a walking robot can’t.

So why is Figure AI building a humanoid? Why is Tesla? Why is Boston Dynamics? Why is Agility Robotics?

The answer is doors.

The world is shaped for us

Look around whatever room you’re in right now. The door handles are at arm height. The light switches are at shoulder height. The counter is at waist height. The stairs are spaced for a human stride. The chair fits a human body. The car has a seat, a pedal, and a wheel, all placed for human limbs.

Every building, every vehicle, every tool, every piece of infrastructure on Earth was designed for a body with two legs, two arms, a torso, and a head. About 1.7 meters tall. About 70 kilograms. Forward-facing eyes. Grasping hands.

A wheeled robot can’t climb stairs. A four-armed robot doesn’t fit in a car. A snake robot can’t turn a doorknob. To operate in the human world, you need a human shape. Not because it’s the most efficient shape. Because it’s the shape the world was built for.

The alternative

The alternative is to rebuild the world. Instead of humanoid robots that can use stairs, build ramps everywhere. Instead of robots that can operate human tools, redesign every tool for robotic use. Instead of robots that can sit in a car, redesign cars.

That would take decades and trillions of dollars. And while you’re redesigning everything, the existing infrastructure still needs to function.

The humanoid form is a shortcut. It’s the cheapest way to make a robot that can operate in existing human environments without modifying those environments. The robot adapts to the world instead of the world adapting to the robot.

The long view

I think this changes over time. Once robots are common enough, new buildings will be designed with robots in mind. Wider doorways. Standardized charging stations. Ramps alongside stairs. The built environment will evolve to accommodate both humans and machines, the same way it evolved to accommodate cars (garages, drive-throughs, highways).

But that’s a 50-year process. In the meantime, the robot that fits the world as it is today is more valuable than the theoretically optimal robot that requires the world to change.

That’s why everyone’s building humanoid. Not because two legs are the best design. Because they’re the most compatible design. And in technology, compatibility beats optimization almost every time.

The answer to “why humanoid?” is the same as the answer to “why QWERTY?” The world locked into a shape, and the shape persists because the cost of changing it exceeds the cost of conforming to it.

Doors. Stairs. Doorknobs. The most mundane features of our built environment are dictating the shape of the most advanced robots we build. I find that funny and a little beautiful.

Related thinking: