On this page
Why the human form?
In science fiction, humanoid robots are either loyal companions like C-3PO or existential threats like the Terminator. Ex Machina offered something more...
In science fiction, humanoid robots are either loyal companions like C-3PO or existential threats like the Terminator. Ex Machina offered something more unsettling: an android intelligent enough to manipulate its creator and walk free. Either way, we keep imagining them in our shape. And now we're building them that way too.
The reason is less philosophical than you'd expect. It comes down to doorknobs.
The Generalist Advantage#
Most robots do one thing well. A robotic arm welds. A warehouse bot carries shelves. A surgical robot cuts with millimeter precision. None of them can do the other's job.
Humanoids are different. Robert Playter, CEO of Boston Dynamics, put it this way in an interview with Lex Fridman: "This is the curse of a general purpose robot, that they're not perfect at any one thing. But they might be able to do a wide variety of things. And that is the goal at the end of the day."[1]
A washing machine cleans clothes more efficiently than any humanoid will. But a humanoid can retrieve the laundry, load the machine, fold the clothes, and put them away. Ask your washing machine to help with gardening.
Could we build something even more versatile, say a robot with three or four arms? Possibly. But as we'll see, the human form has other advantages that make it uniquely suited for mass adoption.
A World Made for Us#
Walk into any building and you'll see the problem. Doors stand six feet tall. Stairs rise seven to eight inches per step. Light switches sit at shoulder height. Doorknobs require an opposable thumb to twist. Every tool in a kitchen, a factory, or an office was designed for creatures with two hands, two legs, and roughly our proportions.
To be precise, the world is optimized for an average-sized adult. If you're a child or in a wheelchair, navigating it is already a challenge. But for a machine built to the same template, everything fits.
We reshaped the world for a machine once before: the automobile. That transformation required decades, trillions of dollars, and the wholesale redesign of cities and landscapes. Doing it again for a rolling drone or a spider-legged robot would mean rebuilding factories, homes, and offices from scratch.
Humanoid robots skip this entirely. They fit through our doors, climb our stairs, and use our tools. The humanoid form is the universal USB connector of the physical world.
This compatibility translates directly into economics. Specialized robots often require expensive environmental modifications. Humanoids don't. But they're also not meant to replace robotic arms or automated guided vehicles. Their role is to fill gaps where human labor is scarce or unavailable.
The Data Advantage#
This generalist capability gets stronger with data. In a CNBC interview in March 2024, Nvidia CEO Jensen Huang explained why: "Robots look like people because... we built the world for ourselves, and so the workstations of a factory, the manufacturing line of a factory, was really created for people. And that's the most important reason. The second most important reason is that we have to teach a robot how to be a productive robot, and you need data for that... We have the most examples of humans moving around as just about any other form of data."[2]
Billions of hours of human movement, captured on video, provide the training data that no other robot form can match. And by working alongside humans, humanoids can collect real-time data on workflows, interactions, and edge cases, improving with every shift.
Gradual Integration#
Humanoids can be introduced slowly, which matters.
Picture a production line with four workers. One calls in sick. A humanoid fills the gap, using the same station and the same tools. No reconfiguration needed. Over time, as confidence grows, humanoids can take the shifts nobody wants, particularly nights. In factories running 24/7, finding workers for the night shift is a persistent problem. Humanoids solve it without burning out the day crew.
And if a robot breaks down, a human can step back in and pick up the same wrench. The tools work both ways. This interchangeability means companies can adopt humanoids without betting everything on them, and workers stay skilled enough to take over when needed.
Telepresence#
There is another advantage unique to the humanoid form: it can serve as an avatar.
A humanoid equipped with cameras, sensors, and haptic feedback could be operated remotely by a human wearing a VR headset and a motion capture suit. The operator sees through the robot's eyes, moves its limbs in real time, and feels resistance when it grips an object.
The science fiction version of this appeared in Surrogates, where humans live their entire lives through robotic substitutes. The real applications are more practical. A humanoid could be deployed inside a contaminated nuclear facility or a chemical spill site, allowing an operator to repair equipment or seal a leak without ever being exposed. The human form matters here because the environment, the tools, and the access points were all designed for human bodies.
Convergent Evolution#
The arguments above rest on verifiable facts. This last one is more speculative, a personal conviction I've held for years, long before I became interested in humanoid robots. I believe the human form is one of nature's most effective designs for embodying intelligence, and I draw this conclusion from convergent evolution.
Convergent evolution describes how unrelated species independently develop similar traits to solve comparable environmental challenges. The process is blind. There is no shared blueprint. Yet the same shapes keep appearing.
A striking example is carcinisation: multiple crustacean species, separated by hundreds of thousands of years of evolution, independently converge toward the crab form, with a flattened body and protective pincers. Hermit crabs, king crabs, and porcelain crabs all arrived at roughly the same design from different starting points. Biologists have documented at least five independent instances of carcinisation. The crab shape, it seems, is not an accident. It is a solution.
The pattern repeats in water. Dolphins, ichthyosaurs, and sharks share torpedo-shaped bodies with fins and dorsal stabilizers, optimized for speed through water. They evolved these forms independently, from completely different origins: mammals, marine reptiles, and fish. The ichthyosaur never coexisted with the dolphin, yet it looks remarkably similar. Flight tells the same story. Birds, bats, and insects evolved wings through entirely separate mechanisms, yet all three converged on the same basic principle: a lightweight surface generating lift by moving through air.
Even at the level of individual organs, the pattern holds. The eyes of octopuses and humans, separated by more than 500 million years of evolution, arrived at the same solution: lenses, retinas, and irises. Nature built the camera eye twice, independently, because it works.
Now consider what it takes to be an intelligent species that builds things. You need to move freely across varied terrain. You need limbs free from locomotion to manipulate objects. You need digits precise enough to grip, twist, and assemble. You need an upright posture to support a large brain and give your senses a wide field of awareness.
Bipedal locomotion is the key that unlocks everything else. Walking on two legs freed our ancestors' hands for tool use. Opposable thumbs enabled the fine manipulation that turns a rock into a blade or a wire into a circuit. The upright posture that followed supported a heavier skull, which housed a larger brain. Each trait enabled the next.
Millions of species have existed on Earth. Many are intelligent. Crows use tools. Octopuses solve puzzles. Elephants mourn their dead. But only one species has built cities, written books, and launched satellites. The difference is not intelligence alone. It is intelligence combined with a body that can act on it precisely, at scale, across environments.
If convergent evolution tells us anything, it is that when a physical form solves a problem well, nature arrives at it repeatedly. The crab form keeps appearing because it is an effective design for armored scavenging. The torpedo form keeps appearing because it is an effective design for speed in water. I believe the humanoid form, bipedal, with free hands and opposable thumbs, is an effective design for intelligent manipulation of the physical world. Perhaps the most effective one possible.
This is why humanoid robots may not just be practical. They may be inevitable. Engineers are not choosing the human form out of vanity or science fiction nostalgia. They are converging on a design already tested by billions of years of natural selection, for the same reason that evolution itself converged on it: because it works.
Changing the World Without Changing It#
Humanoid robots will transform how work gets done while leaving the physical world untouched. The doorknobs stay at the same height. The stairs keep their dimensions.
The form is decided. The logic is sound. But logic does not determine speed. How fast will humanoids actually arrive?
References#
[1] Robert Playter interview with Lex Fridman, "Boston Dynamics CEO on Humanoid and Legged Robotics," Lex Fridman Podcast #374, April 28, 2023. https://lexfridman.com/robert-playter/
[2] Jensen Huang interview with Jim Cramer, CNBC "Squawk on the Street," March 19, 2024. https://www.cnbc.com/2024/03/19/cnbc-exclusive-cnbc-transcript-nvidia-founder-and-ceo-jensen-huang-speaks-with-cnbcs-jim-cramer-on-squawk-on-the-street-today.html