Imagine a robot. You are most likely imagining a sleek, metallic, or chromed humanoid. Yet there is nothing like it in our world; robots are made to perform specific tasks, not to imitate Hollywood fantasies. Our cultural understanding of robots is rooted in Karel Czapek's Czech play Rossum Universal Robots, where they were essentially synthetic humans.
It's not easy to imagine a humanoid robot. Engineers are constantly trying to create something like science fiction robots. Recently, an old competitor returned to the field - Toyota introduced the T-HR3. For a humanoid robot, he is quite agile, with a decent grip, several degrees of freedom, and pleasing movements to observe.
This robot works mainly with the help of a remote control system, which allows the user to control the limbs of the robot, applying different pressure on its frame. Completing the picture is a virtual reality headset that allows the user to control the robot's body. Not a word about the price, but it is unlikely that a car with such a complex control system will appear in your list of gifts for the New Year, if you are not a billionaire.
Toyota is familiar with robots firsthand. She introduced the Partner Robots series, whose robots were great at playing instruments, but didn't do much else. Since Leonardo da Vinci achieved similar successes hundreds of years ago, the series, as expected, disappeared. But the T-HR3 is a continuation of these robots, which emerged shortly after ASIMO in 2003.
The Toyota HSR-2 robot is a robotic base on wheels with a simple mechanical arm. It resembles earlier cars from the Willow Garage like the PR-2. The idea of an inexpensive robot that could just roll around and lift stuff, with no ambition to do anything else, has been pretty successful.
So successful that when Robocup, an international robotics competition, was looking for a platform for its @Home robotics competition, HSR-2 was chosen for its ability to work with objects. The HSR-2 was put into an experiment in the care of the elderly and the wounded, but did not find widespread use after five years after its initial release. This suggests that perhaps the most successful multipurpose humanoid robot will not be humanoid at all - and curiously, Toyota now wants to return to more humanoid models a decade after it ditched them.
What's unclear, as is often the case with humanoid robots, is why the T-HR3 is actually needed. Teleoperation bypasses the difficult problem of control, since the machine is controlled by a person remotely. The person takes over all sensory perception, decision-making, planning and manipulation; in fact, the most difficult problems in robotics.
Promotional video:
Maybe autonomy for T-HR3 won't be some kind of feat, sacrificing autonomy, you greatly reduce the use of robots. Since he cannot act on his own, you will have to find a convincing scenario in which you need a telecontrolled robot that is less accurate and much more expensive than a person who does the same. Perhaps one day the robots will gain more autonomy and an improved maneuvering system that will allow humans to intercept control only if necessary if the robot gets stuck.
Toyota said in a press release that it is "a platform with capabilities that enable people to safely assist people in a variety of settings, such as at home, in healthcare facilities, on construction sites, in hazardous locations, and even in outer space." In reality, it is difficult to see that such a robot will become affordable or even useful in a home or hospital environment (unless it is stronger than a person). True, in this case, it cannot be deployed in the disaster zone or in space. These tasks have been discussed in the context of robots for a long time, and few of them have been solved.
Instead, the robot appears to be designed to work with humans. Its design, 1.5 meters high, weighing 75 kilograms, with 32 degrees of freedom, should accurately simulate a human, and not a robot like ATLAS, which is reliable enough to be used in combat conditions.
Toyota focuses on precision engineering - we see a robot balancing on one leg and then demonstrates several yoga poses. Perhaps it really would be more suitable for the elderly than for the military.
The reason humanoid robots remain elusive and unattainable to us is due to a simple cognitive error. We make two bad assumptions. First, we assume that if you build a humanoid robot, make its joints flexible enough, add a little AI, and perhaps program behavior, it can do everything that a human can do. When you see a robot that moves well and appears humanoid, the hardest part seems to be done; of course, this robot can do anything. But in reality, everything is much more complicated.
We also make the opposite assumption: we assume that when we finally have something to replace, they will be perfect copies of our own bodies and brains, which can perform all the functions to which we are accustomed. Perhaps, in fact, the future of robots and AI is similar to their present: specialized algorithms and specialized machines, gradually learning to surpass humans in every imaginable task, while not even close to humans.
It is possible that the T-HR3 is targeting this machine learning concept as a platform for future research. Instead of trying to program an omnipotent robot out of the box, it gradually learns from its human controllers. In this sense, one can see a platform used to explore the limits of what humans can teach robots, simply by allowing them to simulate the sequences of our bodies' movements, in the same way that the exploitation of neural networks defines the limits of learning algorithms from data. No machine can do everything that a human can do, but together they will be far superior to us in everything.
So when you see a new android like the one that Toyota unveiled, feel free to marvel at its technical capabilities and reflect on whether this is a gimmick or a revolutionary step towards human replacement. Just remember that we are already on this path.
Ilya Khel