What do people imagine when they think of a humanoid robot?
Everyone probably notices the appearance first. It looks similar to a human, has legs, arms, let's say a head and a torso. But the most important definition is that it can move autonomously in a human environment and manipulate objects.
How autonomous are today's robots?
We are not at the stage where we would buy a robot, unpack it from the box, start giving it commands and it would carry them out. Industry and science are currently working on this, but we will be there in five years at the earliest. Today, a robot can be autonomous in specific, well-defined tasks that it learns in advance. Typically, it can move independently in the environment and perform tasks that it knows. It can handle small assemblies or come to a warehouse, pick something out of a box and bring it to your table. If a person wanted something different from it that was not previously in the training data, it would not go well.
Where are they currently most useful?
Personally, I am mainly concerned with their deployment in industry. There, it is typically internal logistics, product sorting, order completion or small assembly. Rather simple tasks. It is important to mention that these are mostly pilot projects. It is not possible to talk about robots fully replacing people today or, for example, taking over the full operation of a line. In my opinion, that is not the case anywhere yet.
Do you expect humanoid robots to reach Czech factories?
Pilot projects in the next year or two for sure. There are companies that want to try it, and I think that is a good thing. People need to experience it, find out where the advantages and disadvantages are. Some Czech companies already have robots, but they are struggling with human resources that would be able to do something with a humanoid. Its preparation is a complex task that includes perceiving the robot, its surroundings and then learning specific tasks. For now, it must definitely be someone with a university degree who has experience with robotics.
How much work can a robot do compared to a human?
At the moment, they are slower, less precise and less resistant to unexpected situations. On the other hand, they can help where positions are difficult to fill. For example, in ergonomically demanding work or in dangerous environments. This includes inspections in demanding spaces, such as in mines or petrochemicals.
Are we heading towards robots fully replacing similar professions?
The industry will try to do this. However, just as it was with other industrial robots, humanoids also need service, operators and supervision. So it will not be that there will never be any work. New positions will be created that will over time be more accessible to people without a university education.
At what stage of development are humanoid robots intended for household help?
Similar to industry, this involves pilot testing for specific tasks. They can wipe the dirt off the table or open the dishwasher and put a plate in it. But for anything unexpected, a human operator must come and control the robot remotely. A household is a much more complex environment than, for example, a warehouse, because it is a mess. There are fragile or transparent materials there that are difficult to detect with camera systems. It is a bigger challenge. Does this mean that someone is constantly monitoring the robot's work from a distance and is ready to intervene? I assume that the robot will evaluate itself when it cannot do something. It is then up to the owner whether he wants to call for help from a teleoperator with a remote connection. For example, he has a VR headset (a set that allows you to perceive a virtual environment as a real space, editor's note) and can relatively easily control the robot's movements, its hands and walking. It's a kind of learning loop, because if a humanoid can't do something and a human operator shows it how to do it, it can do it itself later.
Are we at the beginning of a robotics revolution?
Probably not a revolution. I don't think humanoid robots will have as fast an introduction as, for example, large language models, which came a few years ago and changed the form of work and the overall use of artificial intelligence. Here we come across the physics of the robot. It's specific hardware (the physical equipment of the robot, ed. ) that everyone would have to buy. So these are limited resources. There's not as much hardware to train on. There's also not as much data as there was for large language models that had the entire internet at their disposal. It's going to be more gradual and slower.
Will humanoid robots become the embodiment of artificial intelligence?
That's exactly the vision of full autonomy, so that a robot can receive instructions from a human in natural language and be able to distill from that what to do, how to proceed, and then do it. To this extent, it is still limited. But we already use some artificial intelligence models today. Typically, they are the so-called visual-languageaction models that we use to control humanoids, which is a big leap compared to ordinary industrial robots. They can accept text instructions in the form of human speech and perform assigned tasks. However, the success rate is not always 100%.
Do humanoid robots from marketing videos work in a similar way?
It could be. Putting a language model on a robot is not difficult.
It's the same as if people were talking to a chatbot (a program that can communicate with a person using text or voice, ed.), only it's through a humanoid. But I don't think the movements are paired. They're more like pre-learned movements that can be triggered by the chatbot or a human operator. It's not exactly easy for a robot to come up with new gestures on the move.
What prevents the full use of humanoid robots?
If a language model like ChatGPT makes a mistake, relatively nothing much happens. It generates a bad picture or makes a typo. But a humanoid robot will fall or hurt someone. We have to be more careful about this. So that it doesn't hurt anyone or itself. That seems like a big risk to me. At the moment, the industry doesn't take this into account much and is going the way of making robots know everything that is promised. They tend to separate themselves from people so that they don't come into contact with them.
We also monitor situations where owners take robots out in public and let them meet people. Do you think it's safe?
It certainly isn't, and it can only cast a bad light on humanoids. Manufacturers don't declare that robots should work with people. I've already seen several videos where they've injured people. It's the fault of those who do various kicks in the air with them and the like and don't have safety ensured. Robots at the moment are not automatically made to stop in the presence of people or to even detect that someone is nearby.
So don't humanoid robots have safety guarantees yet?
Newly purchased robots don't. We, for example, created them while working with them. You need to make such a superstructure. Ensure that the humanoid is aware of the space around it using sensors, program a reasonable distance or stop when an obstacle or a person appears. I assume that in the future this will be the standard.
Isn't there a risk that the robot will bypass such fuses in the case of full autonomy?
We are striving to ensure that such unpredictable events do not happen. This will certainly be part of the certification in the coming years, which will ensure that humanoids are not dangerous. But you can always simply disconnect the batteries. The robot falls over because it has nothing to drive the motors. There is absolutely no way around that.
Will we commonly encounter humanoid robots on the streets in the Czech Republic?
At the very least, we will certainly encounter them more often than now. For useful tasks. I assume that companies will have a fleet of humanoids for cleaning, for example. Whether everyone will have them at home is a question.
Personally, I would not require it,
and it will not be as cheap as a regular appliance.
Would you buy a robot as a household helper? Maybe I will change my opinion when I have a house, a lot of work around it and little time. But I assume that we will be far enough away that remote connection will not be needed. I wouldn't want to have a device at home that someone would constantly scan my personal space through.
Which country is currently furthest along in the development and deployment of robots?
Definitely China, which is the largest producer of humanoid robots. You can already see them on the streets there, churning ice cream, making coffee, or performing other tasks in various stalls. Research is also progressing in Europe and America. There, development is greatly supported by technological giants, which is a big lead over Europe. The Americans focus primarily on the development of humanoid brains, while the Chinese focus primarily on hardware and the rapid deployment of simple pilots.
Do people in China accept this technology differently than in the Czech Republic?
I would expect that they would. Autonomous cars are also more common there than here. The difference is not in the roads, but in different legislation. In China, they don't take the safety of citizens as seriously. They don't take cybersecurity as seriously, for example. In Europe, on the other hand, they pay attention to consumer protection.
How do Czechs perceive robots?
When you see a model for the first time, the smile prevails, it's something new. Everyone is curious about how it works. But what's interesting is that it changes when the humanoid is 170 centimeters tall and weighs seventy kilograms. There is sometimes a sense of concern, especially when you try to push it and feel its weight. That's a colder reception.