The Robots Are Finally Leaving the Lab

For as long as most of us have been alive, robots have been five or ten years away, maybe even longer.

Even with the held the promise of “someday” the idea was alway there in books, movies, and even cartoons. They were going to cook dinner, fold the laundry, clean the house, care for older relatives, carry our groceries, and handle the jobs humans didn't want to do or could no longer do.

While science fiction gave us entire societies filled with them, reality gave us the Roomba. That may finally be starting to change.

Not because millions of humanoid robots are suddenly about to arrive in our homes. They aren't. Today's most capable robots remain expensive, specialized, and considerably less versatile than the carefully edited demonstrations appearing online sometimes suggest.

But something important is happening. Robots are beginning to move out of research laboratories and demonstrations and into places where they are expected to perform useful work.

This month, Boston Dynamics opened a new robotics training center at Hyundai's manufacturing complex in Georgia, where its humanoid Atlas robot will learn automotive manufacturing tasks. Agility Robotics, meanwhile, has introduced a new generation of its Digit humanoid robot designed to work around people without being permanently separated behind traditional safety cages. 

Factories and warehouses will probably come first. Homes will be much harder. But the larger story isn't simply that engineers are building better robots. It is that artificial intelligence is beginning to acquire a body.

We've Been Waiting a Long Time

Industrial robots aren't new.

Walk through a modern automobile factory and you'll find machines welding, painting, lifting, cutting, and assembling products with extraordinary speed and precision. But these machines aren't usually what people imagine when they hear the word robot.

Traditional industrial robots excel because their worlds are carefully designed around them. Put a robotic arm in exactly the right location. Place an object in exactly the expected position. Program a specific movement. Repeat it thousands of times. Change the environment unexpectedly, and things become much more difficult.

Humans work differently. We can walk into an unfamiliar room, recognize a box, figure out how to pick it up, notice a chair blocking our path, walk around it, find the correct shelf, and put the box down. We barely think about any of this.

For a robot, each step represents an enormous engineering problem.

It has to see the world. Understand what it sees. Decide what to do. Move through the environment. Manipulate an object. Recognize when something has gone wrong. Then adapt.

That is one reason the household robot has remained perpetually just over the horizon.

The physical world is messy.

So What's Different Now?

Several technologies are improving at the same time.

Cameras and other sensors have become better and cheaper. Batteries have improved. Motors and actuators have become more sophisticated. Engineers can train machines inside simulated environments before putting them into the real world.

But one of the most important changes is happening in the robot's brain.

The same advances in artificial intelligence that allow software to recognize images, interpret language, generate responses, and reason across different kinds of information can increasingly be applied to machines interacting with physical environments.

This is sometimes described as physical AI or embodied AI. The basic distinction is fairly simple. A chatbot can understand your request and tell you how to make coffee. An embodied AI system could eventually understand the same request and make the coffee.

That's a much bigger leap than it sounds.

A language model operates primarily in a world of information. A robot operates in a world of gravity, friction, stairs, glassware, children, pets, moving vehicles, oddly shaped objects, and people who don't always behave predictably.

If ChatGPT gives you a bad answer, you can ignore it. If a 150-pound machine makes a bad decision while carrying a heavy object across a room, the consequences can be considerably different.

Giving AI a body dramatically raises both its usefulness and the stakes.

Why Factories Come First

If humanoid robots are improving so quickly, why aren't they headed straight for our homes? Because a factory is much easier.

Factories are engineered environments.

Floors are relatively predictable. Workstations have known locations. Parts arrive in standardized containers. Tasks can be repeated. Access can be controlled. Engineers can study workflows before robots enter them.

That makes manufacturing and logistics ideal training grounds.

Boston Dynamics' new Robotics Metaplant Application Center at Hyundai's Georgia manufacturing facility is designed specifically around that idea. Atlas robots will be trained for automotive manufacturing tasks before broader deployment. 

Agility Robotics has followed a similar path with Digit, focusing on logistics and material-handling jobs rather than promising an immediate general-purpose household servant. Its latest generation is designed for repetitive industrial work while operating more safely alongside people. 

These early jobs aren't particularly glamorous. Move this container. Pick up that part. Carry these materials from one location to another. Repeat. But boring is important.

A robot that performs one useful task reliably for thousands of hours is economically much more interesting than one that performs twenty spectacular tasks once for a promotional video.

That gives us a useful rule for evaluating the flood of robot videos we're likely to see over the next several years:

Don't just ask what the robot can do. Ask how reliably it can do it.

Why Do Robots Need to Look Like Us?

There is another obvious question. Why build humanoid robots at all? Wheels are simpler than legs. Specialized robotic arms can be better than humanlike arms.

A machine designed specifically to move boxes doesn't need a head, torso, hands, and two feet. And in many situations, specialized robots will continue to make much more sense.

But humanoid robots have one enormous advantage: The world has already been built for humans.

• Think about your surroundings.

• Doors are positioned for human hands.

• Stairs are designed for human legs.

• Shelves are built around human reach.

• Tools have handles made for human fingers.

• Factory workstations are built at human height.

• Cars have steering wheels and pedals positioned for human bodies.

Warehouses, kitchens, offices, hospitals, stores, and homes have all been constructed around roughly the same physical template. Us.

Instead of redesigning the entire environment for robots, engineers can try to build robots capable of operating inside the environment we already have. That doesn't mean every future robot will look human. But it explains why companies are investing so heavily in machines that roughly share our physical form.

A Robot That Works Beside You Is Different

There is another transition happening that may ultimately matter more than whether a robot has two legs.

For decades, many industrial robots have operated behind barriers. That's partly because powerful machines moving quickly through predictable motions can be extremely dangerous if a person unexpectedly gets in their way.

The next generation of robots is increasingly being designed around collaboration.

Agility's latest Digit, for example, incorporates safety systems intended to allow the robot to work in closer proximity to people, slowing or stopping when necessary rather than requiring permanent separation behind a cage. 

That sounds like a small engineering improvement. It isn't. A robot that must operate inside a carefully isolated zone is essentially another piece of factory machinery.

A robot capable of safely sharing human spaces becomes something else.

It can potentially work in warehouses. Stores. Hospitals. Hotels. Airports. Eventually, perhaps, homes.

The ability to operate safely around ordinary people is one of the thresholds robots have to cross before they become part of everyday life.

Robots Probably Won't Replace “Jobs” All at Once

This is also where the conversation inevitably turns to work. Will robots take people's jobs?

It's an understandable question, but probably not the most useful one.

A better question is: Which tasks will robots become good enough and cheap enough to perform?

Most jobs are bundles of tasks.

A warehouse employee might move merchandise, inspect packages, solve unexpected problems, interact with coworkers, operate equipment, organize inventory, and respond when something goes wrong.

A robot doesn't have to replace that entire person to change the job. It may simply take over the repetitive movement of containers. Likewise, a hospital robot doesn't have to become a nurse to transport supplies between floors.

A hotel robot doesn't have to replace the front-desk staff to move linens.

A manufacturing robot doesn't have to replace an assembly worker to handle one physically demanding part of the workflow.

The earliest opportunities are likely to involve tasks that are some combination of:

repetitive, physically demanding, dangerous, structured, or difficult to staff.

And as robots spread, other human jobs will grow around them—installing, maintaining, supervising, training, troubleshooting, auditing, and integrating machines into existing workplaces.

The interesting question isn't simply how many jobs disappear.

It's how jobs get reorganized when some physical tasks can be delegated to machines.

Your House Is a Robot's Nightmare

Eventually, of course, most people aren't going to care about robots because one can move automobile parts across a factory.

They're going to ask: When can one fold my laundry?

That's where expectations need some adjustment. Your home may feel perfectly ordinary to you. To a robot, it is chaos. A child's backpack is lying where it wasn't yesterday. The dog runs through the hallway. Someone left a glass near the edge of the counter. A sock is underneath the chair. The dishwasher contains twenty objects of different sizes, shapes, materials, and fragility.

You tell the robot: “Clean up the kitchen.”

Think about how much knowledge is hidden inside that seemingly simple instruction.

• Which objects are garbage?

• Which go in cabinets?

• Where does each object belong?

• Is that pan still hot?

• Should the unopened package on the counter be put away?

• Is the half-full glass someone's drink or something to clean up?

• Can this container go in the dishwasher?

• And what happens when a toddler walks into the room halfway through?

Humans solve hundreds of these tiny problems almost unconsciously.

A general-purpose household robot has to solve them deliberately—and safely.

That's why impressive humanoid demonstrations shouldn't be confused with widespread household readiness. As of early September, only a small number of home-oriented humanoid robots were even being offered for purchase or reservation, and consumer deployment remained extremely limited. 

Factories are structured. Homes are improvisation. That gap is enormous.

Robots Were Waiting for Better Brains

And yet this is precisely why today's AI advances matter so much.

For years, robotics faced a difficult tradeoff. Machines could be extraordinarily precise when engineers programmed exactly what they should do. Or they could become more flexible—but much less reliable.

Modern AI offers another possibility.

Instead of programming every possible situation individually, developers are working toward systems that can understand instructions, perceive their surroundings, learn from demonstrations, and apply what they've learned to situations they haven't encountered in exactly the same way before.

In other words, the goal is increasingly not: Program the robot to perform this motion.

It is: Tell the robot what you want accomplished.

That is a profound difference.

The robot still needs an enormous amount of engineering underneath it. It still needs motors, sensors, control systems, safety mechanisms, and training. But increasingly, the intelligence coordinating those systems may become more general.

That's why some people in the robotics industry talk about searching for a “ChatGPT moment”—a point at which improvements in AI suddenly make robots useful across a much wider range of tasks rather than one carefully programmed activity at a time. 

Whether that moment arrives next year or much later remains uncertain. But the direction is becoming clearer.

What Happens When AI Can Touch the World?

The past few years introduced millions of people to artificial intelligence through a text box.

We type. AI responds. That's powerful, but there is still a boundary between the digital system and the physical world.

Robotics begins to erase that boundary.

• An AI agent might book your flight.

• A physical AI system could someday carry your suitcase.

• An AI assistant might remind an older person to take medication.

• A robot could potentially retrieve it.

• An AI system can tell a warehouse manager how inventory should be moved.

• A robot can actually move it.

That is why the current wave of robotics matters even if you aren't particularly interested in humanoid machines. It's another stage in the same technological transition we've already been watching.

First, computers calculated. Then they connected. Then they learned to recognize and generate information. Now we're trying to teach them to act.

The Robots May Arrive Quietly

The first successful humanoid robots probably won't look much like the robot future science fiction promised us.

• They may spend their days moving boxes.

• Carrying automobile parts.

• Loading carts.

• Delivering supplies.

• Performing monotonous tasks inside buildings most of us never enter.

• That's probably a good thing.

• Transformative technologies rarely arrive everywhere at once.

• They start where they solve a specific problem well enough to justify their cost.

• Then they improve.

• They become cheaper.

• They move into adjacent jobs.

Eventually, something that once seemed extraordinary becomes infrastructure.

For decades, robots mostly lived behind safety cages, inside research laboratories, or in our imagination. Artificial intelligence may finally be giving them enough adaptability to venture beyond those boundaries.

The household robot that cooks dinner, folds the laundry, and cleans the kitchen isn't here yet. But for the first time in a long time, the path from the factory floor to everyday life is becoming easier to see.

The question is no longer simply whether useful humanoid robots can be built.

It's how long it takes before seeing one at work no longer feels futuristic at all.

Stay ahead of the curve with the latest in Web3 culture and innovation. Subscribe to Hashed Out for exclusive insights, case studies, and deep dives into the decentralized future.

Help Grow Hashed Out And Get Rewarded With Premium Content & Merchandise

If you believe in a more open, fair internet — help us build it, one reader at a time.

Web3 adoption starts with curiosity. Share Hashed Out with someone who’s ready to explore.

You’re not just sharing a newsletter — you’re inviting someone into the future of digital life.

Refer 3 friends and unlock premium content. The more you share the more rewards you unlock, including Hashed Out mugs or tote bags, and exclusive community memberships.

Do you find this article useful?

• CHECK OUT OUR PREVIOUS EDITIONS