Do humanoids dream of changing into human?

Tales of human-like dolls craving to turn out to be actual individuals flip up in all places. Pinocchio desires to be an actual boy. The robotic little one in Spielberg’s A.I. desires to be cherished like a human son. The story retains getting retold as a result of individuals assume the trajectory is clear. Construct one thing that appears human, maintain enhancing it, and in the future the copy turns into indistinguishable from the unique.

What’s taking place on the bottom is stranger than that. At CES 2026, Boston Dynamics’ Atlas demonstrated wrists that bent backward and a torso that spun a full 180 levels. Elsewhere, humanoid robots are starting to diverge in much more hanging methods. Some can swap their very own batteries by reaching each arms behind their backs. Others stroll on reverse-jointed legs. The human silhouette remains to be there, however the actions inside it have gone some other place solely.

There’s an apparent objection right here. Hasn’t copying nature labored earlier than? Typically. Gecko toe pads gave engineers the thought for dry adhesives. Sharkskin texture confirmed up in aggressive swimsuits. However in each instances, engineers borrowed the physics beneath, not the form. Those who tried to repeat pure types wholesale often hit a wall. 

For hundreds of years, individuals tried to construct ornithopters that flapped like birds, however none grew to become a sensible path to human flight. The Wright brothers received off the bottom not as a result of they merely imitated, however as a result of they moved past flapping and targeted on the ideas of elevate and management.

If evolution has spent thousands and thousands of years refining a design, why don’t engineers simply copy it? That query went to the Hubo Lab at KAIST. The lab constructed HUBO, the robotic that gained the 2015 DARPA Robotics Problem, and at the moment it’s led by Prof. Park Hae-won. His group’s latest work offers a way of the vary. Humanoid legs that dash at 12.6 kilometers per hour. A quadruped robotic that walks straight up vertical partitions. A one-legged hopper that launches into mid-air somersaults and lands on the identical leg.

Mimicking nature just isn’t at all times the best reply.

At 12.6 kilometers per hour, an individual has to interrupt right into a run. A robotic constructed by Prof. Park Hae-won’s group at KAIST can dash at that velocity on two legs. It glides by way of motions that seem like Michael Jackson’s moonwalk and picks its approach over tough terrain with a duck-like waddle. 

One place to start out is biology. Roboticists have been borrowing nature’s methods for many years. Prof. Park’s robots do seem like they arrive from that custom. However he works the opposite approach round. As a substitute of learning an animal to construct one, he picks an issue and builds a machine to unravel it.

“For those who’re creating know-how for high-speed motion, wheels could be an environment friendly alternative,” Prof. Park mentioned. “There’s no must mimic the movement of a cheetah.”

A automotive on wheels outruns a cheetah. Evolution by no means got down to construct the quickest runner. It constructed the one almost certainly to outlive.

“Learning pure organisms offers us a way of the extent of efficiency that may be reached when one thing is properly designed,” Prof. Park mentioned. “It serves as a helpful reference for setting path throughout analysis and growth.” He added “It’s necessary to view nature as one reference level. Moderately than replicating it instantly, it’s extra applicable to make use of it as a supply of concepts.”

Humanoids face the identical query. A human physique runs on muscle tissue, tendons, and chemical vitality. A robotic runs on steel frames, motors, and electrical energy. To repeat human motion faithfully you’d want synthetic muscle tissue, however motors nonetheless are inclined to outperform commercially accessible synthetic muscle tissue in lots of sensible metrics. So why handicap a robotic by forcing it to maneuver like a physique it doesn’t have?

MARVEL, a quadruped robotic from Prof. Park’s lab, was designed for grimmer work. Researchers wished a robotic that might transfer freely throughout the metal constructions of shipyards, bridges, and huge storage tanks. Locations the place upkeep crews threat deadly falls.

Gecko toes or insect claws may sound like the best mannequin for a wall-climbing robotic. However actual industrial metal is rusted, layered in outdated paint, and caked with grime. Gecko-style adhesion would seemingly battle to carry heavy tools on surfaces like that.

As a substitute, Researchers constructed MARVEL with electro-permanent magnets in its toes. Standard electromagnets drain energy constantly to remain on. Electro-permanent magnets work in another way. A quick electrical pulse rearranges the interior alignment of the magnet’s poles, switching the grip on or off. MARVEL’s toes lock and launch in about 5 milliseconds.

As soon as the magnets interact, the wall itself turns into the robotic’s floor. Three legs keep anchored whereas the fourth steps ahead. MARVEL travels at 0.7 meters per second on vertical partitions and at 0.5 meters per second whereas hanging the wrong way up from a ceiling. Its adhesive drive reaches almost 54 kilograms, which is sufficient to carry not simply its personal weight but additionally heavy instruments.

“For those who method a shipyard robotic from a biomimetic perspective, you may conclude that it ought to resemble a human employee and deal with instruments the identical approach,” Prof. Park mentioned. “In the end, what issues is designing a system that matches the working setting and the duty at hand.”

AI alone can’t construct an ideal robotic.

Designing the physique is simply half the issue. AI and reinforcement studying have modified how robots be taught to maneuver, however what works in simulation nonetheless has to carry up on actual {hardware}.

Prof. Park’s group trains its robots by way of reinforcement studying. The AI controls the robotic’s physique and figures out the best way to stroll by trial and error, falling and getting again up the best way a toddler does. Doing that 1000’s of instances on actual {hardware} would take ceaselessly. So researchers prepare in simulation as a substitute.

Contained in the simulation, Prof. Park’s group runs roughly 400 copies of the identical robotic without delay. Every copy falls and recovers underneath totally different situations, and what all of them be taught feeds right into a single AI community in actual time. Time itself could be compressed. What would take a few yr of bodily observe matches into roughly 4 hours on a high-performance laptop. Prof. Park mentioned half a day of reinforcement studying is sufficient to get a robotic strolling.

The catch is {that a} robotic educated in simulation doesn’t at all times survive contact with actuality. A robotic that tumbles like a gymnast on display screen can lose its steadiness and topple the second it’s positioned on an actual flooring. Roboticists name this the sim-to-real hole. Simulations can’t seize each wrinkle of real-world physics, and the variations are sufficient to throw off an AI that realized in an easier world. Closing that hole is the place the KAIST group’s {hardware} experience is available in.

One method Researchers took was to make the true robotic behave extra like its simulated twin. An enormous purpose AI struggles to regulate a bodily robotic is friction within the joints. Standard robots use off-the-shelf reducers with excessive gear ratios to amplify motor output. That provides the robotic highly effective drive. On the similar time, inside friction makes every little thing stiff, like pedaling a bicycle caught in excessive gear.

“In a gear system with a excessive discount ratio, it’s very laborious to drive it to show from the surface,” Prof. Park mentioned. “For those who connect a linkage and strike it with a hammer, the resistance is so intense that the gear enamel may shatter.”

Most simulations don’t account properly for that friction. An AI that realized to stroll in a near-frictionless digital world loses its steadiness the second it hits the stiff resistance of an actual joint. So Prof. Park’s group constructed its personal actuator that lower the gear ratio to roughly one-tenth of standard ranges whereas boosting the motor’s personal output. It’s a quasi-direct drive design, an idea first proposed at MIT. Much less friction within the {hardware} meant the true robotic moved extra just like the simulated one. After the adjustment, AI’s coaching truly carried over.

KAIST group additionally labored the issue from the opposite path. As a substitute of constructing the {hardware} match the simulation, they made the simulation match the {hardware}. As a result of Prof. Park’s group designed and constructed its personal motors, they’d detailed knowledge on how these motors truly behave.

That knowledge issues. Most simulations assume torque stays the identical irrespective of how briskly the motor spins. Actual motors don’t work that approach. Spin sooner, accessible torque drops. Decelerate, accessible torque climbs. Coaching an AI on the simplified model will drive it to push the {hardware} past its limits. Prof. Park’s group fed their precise torque-limit curves into the coaching, so the AI realized the place the motor’s ceiling was and stayed underneath it.

The place all of this comes collectively is KAIST’s hopping robotic. The entire machine is one leg. No arms, no second foot to catch itself. That type of steadiness downside is brutal to unravel. In the meanwhile Prof. Park had already gotten quadruped leg robotic strolling to work. As a substitute of shifting to 2 legs subsequent, he went straight to 1. As a result of If the algorithm can deal with the toughest case first, then two legs gained’t be an issue.

KAIST Humanoid v0.5

Researchers loaded every little thing about the true robotic into the simulation. Its shifting heart of gravity, its inertia, and the bodily limits of its actuators. From there they ran almost the identical reinforcement studying algorithm they’d used for the quadruped. The AI discovered the best way to steadiness on one leg. It began leaping. Earlier than lengthy it was doing mid-air somersaults, touchdown cleanly every time.

“Constructing the hopping robotic confirmed that our reinforcement studying algorithm and {hardware} design could be utilized underneath a variety of situations,” Prof. Park mentioned. “It gave us a possibility to discover how our motor know-how and reinforcement studying methods may prolong to the event of robots in many various types.”

Prof. Park doesn’t purchase the concept software program can resolve every little thing. He’s watched junior researchers spend days debugging code when the true downside was a unfastened screw or a damaged solder joint. When a robotic gained’t stroll, individuals attain for the algorithm first. They tweak the parameters, rerun the simulations, rewrite the management logic. In the meantime the precise fault is sitting proper there within the {hardware}. No quantity of code will tighten a screw. {Hardware} data isn’t going away simply because AI received good.

“Irrespective of how refined the management know-how, there are limits to what could be achieved if the {hardware} can’t sustain,” Prof. Park mentioned. “In robotic growth, management and {hardware} are each essential. Neither could be thought-about in isolation.”

Can humanoid robots turn out to be a part of our on a regular basis lives?

The cash pouring into humanoid robots proper now’s staggering. However loads of applied sciences have appeared simply as promising and gone nowhere. Honda spent over 20 years on ASIMO earlier than quietly retiring it. A robotic that walks throughout a stage at a commerce present just isn’t the identical factor as a robotic that survives a shift on a manufacturing unit flooring.

Prof. Park’s humanoid is being constructed for the manufacturing unit flooring. The goal payload is 25 kilograms or extra. Most humanoids available on the market high out properly under that. He selected that quantity due to the place South Korea is correct now. The nation runs one of many world’s largest manufacturing sectors, however the workforce is graying quick. Younger individuals aren’t lining up for welding jobs or assembly-line shifts. The slack is being picked up by older expert employees and overseas laborers, and there aren’t sufficient of both. A robotic that may solely carry gentle objects is ineffective in that setting. The quasi-direct drive actuators and customized motors his researchers have been constructing exist for precisely this type of work.

The manufacturing unit flooring isn’t the one doable market, although. Prof. Park introduced up drones. For many years solely the navy and some infrastructure inspectors bothered with them. Then YouTube creators began wanting aerial pictures and went in search of one thing that might fly a digital camera. Drone firms shipped an affordable quadcopter with a good digital camera mount. Inside a number of years a client drone business had grown up round a necessity that hardly existed earlier than. Prof. Park thinks humanoids may go the identical approach. The use that truly drives adoption is perhaps one no one within the business has imagined but.

On the shut of the interview Prof. Park mentioned, “I imagine robots ought to complement individuals, not compete with them. My hope is that robots will in the end be used to counterpoint individuals’s lives and free them to pursue extra fulfilling work.”

The story was produced in partnership with our colleagues at Well-liked Science Korea.

 

2025 PopSci Better of What’s New

The 50 most necessary improvements of the yr