Shadowing Practice: The Sensor That Changed Robotics Forever - Learn English Speaking with Video

This is a bumper. A strip of plastic on every modern robot vacuum since two thousand and two.

The cheapest sensor possible - it's there so the vacuum knows it has bumped into something.

In 2002, the Roomba came out with a bumper just like this one. Here's the paradox: the people who built the Roomba were a group of AI researchers from MIT. Some of the most influential figures in robotics and AI at the time. And their entire ideology was built around this bumper. Which actually created a new category of household devices and an entire market around it.

But by then, robot vacuums already existed on the market. And they were a lot smarter.

They had machine vision, they mapped rooms, they did a lot more. But it was the Roomba with its bumper that effectively pushed all of them off the market and set the new standard.

This is a design and conceptual decision that has shaped the modern world more than any of today's humanoid robots. And the same decision explains why projects like Optimus and Atlas are dead ends.

I'm going to tell you the philosophy and the history behind this piece. Why it killed all of its competitors. And how it happened that its own creators - iRobot and Roomba - drifted away from their own philosophy and ended up losing the market.

And who took the market from them, and why those companies are the real successors to the concept.

This video is made in partnership with Ecovacs. More on that later.

By the way, meet Yuiko. My fourth robot vacuum. I've never named a vacuum before.

But my little son keeps petting it and saying "good vacuum," and his way of humanizing it turned out to be contagious. One in four Americans my age has a robot vacuum at home. And a lot of them give it a name. Most often it's Rosie - after the robot maid from The Jetsons - or Bobby, a nod to the word "robot." If you look at the list of inanimate things we name, it's mostly soft toys and house plants. Sometimes cars. Sometimes musical instruments.

A PlayStation 5, a phone, or a kettle don't get names. A robot vacuum does. Even though, on the surface, it has more in common with a kettle than with a cat.

I want to treat this as a kind of signal. A name is a marker that an object has crossed from the category of "appliance" into the category of "something like a creature." And for some reason, the robot vacuum is the device that's making this transition at scale.

What makes a piece of plastic on wheels alive enough to deserve a name? Unlike Japanese robot dogs or those uncomfortably realistic robots from Boston Dynamics, it doesn't imitate living movement or emotion. So what is it about this thing that makes us humanize it? The first answer that comes to mind is movement. It moves around, that's why. That's close to the truth, but the real question isn't that it moves, it's how it moves. And that's exactly what's defined by this same bumper - the collision sensor.

Which is where the entire philosophy of the robot's AI lives, according to those same MIT researchers, Brooks and Jones. This philosophy solved two problems. One obvious - cleaning quality. And one hidden - distrust of robots.

I want to spend a moment on the second one. We're actually living through a period of growing fear of robots and AI. And meanwhile, for about twenty years now, an entire segment of smart, autonomous home devices has been growing. Including ones with cameras. Including ones that have already had data leak scandals. But the category of robot vacuums exists almost outside this discourse about robots and AI in our lives. And yet at the start of this technology, before the Roomba, the concerns were sharper. Mainly because automatic cleaning was perceived as an intelligence problem. It made sense - in order to clean properly, the device should understand the space and make decisions inside it. The first notable attempts went down this path. Back in 1997, on a BBC show called "Tomorrow's World," they showed a robot vacuum from Electrolux that already looked like modern robots from the outside.

It was a very complex device, and it took five more years to reach the market. It cost $1,500 and it cleaned poorly. And the name was Trilobite. I get it - a flat round robot makes you think of one. But releasing a household robot with complex AI systems and naming its design after an ancient, kind of creepy arthropod is a questionable choice. At the same time, Dyson was working on their own model. They couldn't get the cost below $3,000 just to manufacture, which seemed too expensive to release, so they shut the project down. I want to show you two more robots. HAL 9000 and R2-D2. Yes, one's the antagonist and the other is the trusty friend, just by virtue of the plot. But let's try to see what the authors used to underline these characters.

Kubrick made HAL silent and quiet. Clearly something was always processing inside him, but you couldn't see any of it. A completely opaque being, smart, with no feedback. Almost a bodiless ghost with access to everything. And what made HAL terrifying was exactly this opacity.

Lucas made R2-D2 smart too, but to keep him from being scary, he made him completely transparent.

All his emotions came through sounds. He had clear physical limits defined by his body. He gave constant feedback, even when he was alone, which made him feel understandable and predictable.

Early prototypes of smart robot vacuums looked terrifying for the same reason. They'd roll into the middle of a room and start running calculations. Whatever they were thinking and deciding stayed opaque to the human watching, and then they'd start working. Most of the time they'd get stuck against a wall, because their machine vision had mistaken sunlight on the wall for a continuation of the room. So naturally, Electrolux and Dyson tried to fix this by making their algorithms more complex, adding more sensors, building more elaborate models of space. The surprising thing is that a real robot vacuum - one that worked, didn't scare people, and could be cheaper - existed as a prototype for years. The creator even tried to sell the idea to companies, but there was no demand.

In 1989, a graduate student named Joe Jones built a prototype cleaning robot out of LEGO.

Under the supervision of Rodney Brooks, one of the most prominent voices in robotics at the time.

Brooks was the leader of an AI movement whose central ideas in the 80s sounded almost heretical. For a robot to work in the real world, it doesn't need artificial intelligence in the sense everyone else understood it. It needs a body that can interact with the world directly.

In 1990, he published an article called "Elephants don't play chess." The thesis: intelligence isn't the ability to solve logical problems. It's the ability to survive in a dynamic world. An elephant is smarter than any chess computer, because the elephant can find water in the savanna.

From this philosophy, Brooks derived a central principle - embodiment. The robot's body should be its main sensor. Not a camera, not a laser, not a processor - the construction itself should give the robot information about the world through direct physical contact.

The best collision sensor for hitting an object is the thing that physically runs into it. A bumper.

And it was the bumper that solved both problems I mentioned earlier. The cleaning quality problem - real, working navigation without a complex spatial model. The trust problem - every action the robot takes has a visible cause. The robot bumps into a wall, you see it bump into a wall.

This robot looked more like a slightly clumsy baby animal than a smart, secretive machine.

Beyond embodiment, Brooks had three more principles. The world is the best model of the world: don't build a map if you can just drive through. Subsumption: behavior isn't one complex program, it's several simple layers, each with its own job. Fast, cheap, and out of control: don't try to make it perfect, make it rough and iterate. And one more important detail from the same "body as sensor" logic. The dirt sensor in a Roomba isn't a camera. It's a piezo membrane inside the body that physically hears dust particles hitting the plastic. The Roomba literally hears the dirt.

In 2002, the Roomba came out, built on these principles. From iRobot, the company founded by Brooks and Jones. It cost $200. It cleaned much, much better. And it stirred emotion.

The arrival of the Roomba was, in effect, something like skeuomorphism in graphical interfaces. A bridge that breaks through perception barriers, a way for the world to get used to living with robots. We still haven't moved past this stage, because the progress in consumer robotics hasn't really spread into other areas yet.

Within robot vacuums themselves, things have moved far ahead. And the surprising thing is that iRobot, the founders of the concept, at some point veered off their own path. They started saying the same thing Tesla is saying right now: if humans see space visually, then artificial intelligence should orient itself the same way. Machine vision is a complex, expensive, and most importantly slow technology. And iRobot, having taken a near-monopolistic position on the market, started focusing on exactly that kind of research. It was already obvious that navigation quality was good enough. And it got significantly better when competitors started using lidar - a cheap, fast way to know where obstacles are. Much cheaper than processing camera images.

The competitors moved on to the next problems. How to do real wet cleaning. How to make robots more autonomous. And they just walked past iRobot, who got stuck in their own ambitions.

Today, the leader of innovation in this category is Ecovacs. For full transparency, let me explain the nature of this integration, because it's not a regular ad spot. It's a continuation of the topic of this video, illustrated through a product. I've been interested in robot vacuums - their history, their design - for a long time, and I was looking for someone who could sponsor this video. The only companies and products that could fit were ones I could honestly tell the whole story through. Ecovacs's new DEEBOT X9S Pro Omni is a perfect fit.

Because the main pain points of older robot vacuums are solved here. And for our video, what's interesting is to look at these design decisions as a continuation of the embodiment philosophy and the development of consumer robotics in general, rather than as a list of features.

My previous robot vacuum was also a mopping one. I didn't upgrade it for years, because I didn't see much point. It already did regular cleaning well enough, and the times when it didn't were usually because of something the human had to handle - washing and drying the cloth, swapping the water, cleaning out debris. I haven't been able to imagine my life without a robot vacuum for a long time. We have a dog at home, a kid, and even a carpet. My wife and I are both from Asian backgrounds, so we walk around the house completely barefoot, which makes us even more sensitive to clean floors. So even though the old robot was constantly mopping, we still ended up mopping by hand sometimes. But what really surprised me was that after the first cleaning with the new vacuum, the water in the tank was very dirty - almost black. On the second pass, noticeably cleaner. By the third, mostly clear.

There are three reasons the cleaning quality went up so much with this robot. First, the kind of mopping itself. It's a real roller that doesn't just spin - it presses down on the floor. The body is heavy enough to apply more pressure where it's needed. And of course, a rotating mechanism is in another league compared to dragging a cloth across the floor. Second, it can mop places the old robot couldn't reach. The X9S Pro Omni extends its brush sideways, all the way out. So it goes right up against the wall and doesn't leave gaps. And third, I configured it to automatically return to the dock at intervals and rinse the brush. That helps a lot - instead of just smearing the dirt evenly across the surface so you don't see it, it actually removes it. Plus, while it's at the dock dumping the dustbin and rinsing the brushes, it also charges - 10% in 3 minutes.

So the only thing left for me to do is dump the dirty water and refill with clean. That's exactly the weak point old robot vacuums had. So much annoying maintenance. I have long hair, and it's everywhere in the apartment, mixed with dog fur. With the old robot, at least once a week I'd be cleaning hair and fur tangled around the brush. With this one, somehow, the hair and fur don't get tangled at all. I'm also not afraid to start it in an unprepared room. Furniture, toys, everything stays where it is, and that doesn't stop the robot from doing a light daily clean. Not once has it caught a cable. For a deep clean, sure, you should clear the floor, but the philosophy shifts from rare deep cleans to constant background autonomous maintenance. I noticed a behavioral shift in myself. I used to default all my vacuums to maximum power and maximum water flow. With the X9S, I ended up on medium settings, because that's more than enough for the apartment to stay consistently clean.

It really is a meaningful step in automated cleaning. After several months of using it, I can say it requires almost no attention, and the apartment is noticeably cleaner.

But that's the personal experience. As for the concept, it's all happening through Brooks's principles. Through direct, understandable solutions to problems. Need to mop better?

Make the body heavier so it presses down on the mop. That's a clear example. For autonomy, the systems do get a lot more complex. And of course, modern robot vacuums have many more functions than even the over-engineered Trilobite of 2001. But that's just a different generation, solving different problems, where this kind of investment becomes justified. The old models tried to be brain-first, and that ended in failure. The modern ones grow brains inside a working body, in service of making the body work better. Remember I mentioned Tesla earlier. We've been promised self-driving cars for years, but Tesla's autonomy still loses to the likes of Waymo. And the difference between them is very similar. Tesla's AI philosophy leans on imitating humans, while all the lidar companies go down the path of simplification. Which, as you can see, has its benefits. The same can be said about robots, which - surprise - Tesla also makes. Imitating the human body and the physics of human movement has its pluses, and it opens up possibilities that wheeled drones don't have.

Right now, Boston Dynamics robots are being used for dangerous work in extreme conditions.

It all looks futuristic. But to me, this is more like retrofuturism. Bright images that hit our psychology directly. But this isn't a direct path to solving specific problems. It's an attempt to build a universal system that can solve everything.

Brooks's philosophy is different. We start by accepting the limits of natural perception, of animals and of humans. And from there, with no extra grandeur or pretension, we look for the shortest path. And the practice shows that this path works.

Humanoid robots are something out of the modernist era, where humans seem like nature's perfect creation. But in reality, evolution has saddled us with so many problems and complications, things we have to deal with constantly, that imitating the result of that evolution looks, at the very least, like a strange choice. On the other hand, evolution led to us being here and functioning. So the answer is to imitate evolution itself, not its end product.

And it's exactly because Brooks set up that frame that we already have robot vacuums, delivery robots, drones, and even autonomous assembly lines.

It's just that none of them get this much attention, because they quickly become the new norm and a natural part of our lives. And we don't see them as an existential threat.

We see them as a friendly little helper like Yuiko, my Ecovacs X9S Pro Omni, who occasionally needs to bump into a chair leg to know it's there. And here I'd like to draw a parallel not just to robots, but to myself. Building overly complex models of the world is sometimes fun, but more often, just going and bumping into things, taking it lightly and moving on, is more effective, simpler, and cheaper. I put a lot into this video. If you enjoyed it, that means you're into digging into the surprising stories behind ordinary-looking things, alongside me. Like the Like button - there's much more to it than it seems. Ancient instincts, the American military, and 9/11 are all part of the story. Definitely check it out.

And thanks to ECOVACS for sponsoring this video. See you. Bye.

What is the Shadowing Technique?

Shadowing is a science-backed language learning technique originally developed for professional interpreter training and popularized by polyglot Dr. Alexander Arguelles. The method is simple but powerful: you listen to native English audio and immediately repeat it out loud — like a shadow following the speaker with just a 1–2 second delay. Unlike passive listening or grammar drills, shadowing forces your brain and mouth muscles to simultaneously process and reproduce real speech patterns. Research shows it significantly improves pronunciation accuracy, intonation, rhythm, connected speech, listening comprehension, and speaking fluency — making it one of the most effective methods for IELTS Speaking preparation and real-world English communication.

Shadowing Practice: The Sensor That Changed Robotics Forever - Learn English Speaking with Video

Why practice speaking with this video?

Grammar & Expressions in Context

Common Pronunciation Traps

What is the Shadowing Technique?