Why humans are superior to robots
Machines may be able to perform special movements with outstanding speed or precision, and individual components may be more robust and stronger. But a comprehensive analysis now shows that when it comes to managing our everyday world, we are unrivalled - at least for the time being.
Machines have faster and stronger components, and they master individual movement sequences perfectly. But humans are more versatile and therefore better equipped for a varied environment. Even the 27 most powerful robots that can be compared with us may beat us in specific disciplines, but cannot outperform us in general. This has now been shown by a comprehensive analysis carried out by Robert Riener, Professor of Sensorimotor Systems at the Swiss Federal Institute of Technology in Zurich, together with two colleagues.
The research, published in November 2023 in the journal "Frontiers in Robotics and AI", was prompted by the enormous technological progress made in recent decades. We have long had industrial robots working for us on conveyor belts and in other repetitive production steps. They can do much more than humans can, both with heavy workpieces and with precision work. Robots are now taking on more and more tasks beyond production halls and warehouses. Robot hoovers are rolling through countless living rooms, delivery robots are increasingly bringing goods or food, and more complex, more humanoid machines are already becoming part of everyday life, for example as assistants in care.
Such developments "fuel prejudices and even fears in society", write Riener and his colleagues with reference to studies on technophobia. There is still no reason to fear dystopian scenarios like those from science fiction films such as "Terminator", in which superior machines make humans obsolete in the foreseeable future. But in view of the rapid progress, scientists are asking: "Where are we today? Can robots already replace humans in typical human activities?" In the systematic search for answers, criteria were first needed to make a meaningful comparison possible in the first place.
Any robotic arm can be made more powerful and more enduring by increasing its size and power source. At some point, however, it will no longer fit through any door. Machines can move faster and more efficiently on wheels than humans on legs, but fail when faced with an obstacle such as a staircase. In order to operate in an environment created by humans for humans, robots must therefore fulfil certain criteria: Among other things, they need minimum and maximum masses, should be able to grasp things and also have legs (although this may well be more than two) on which they can overcome barriers.
Thus, out of thousands of existing robots, only 27 were ultimately left that fulfil all the criteria (in cases of doubt with simple upgrades such as arm extensions). The 13 most recent of these have been presented to the public over the last ten years. These include "Spot" from the US robotics company Boston Dynamics, which has been commercially available since 2020 and has become somewhat famous thanks to spectacular promotional videos on the internet.
More than the sum of its parts
The researchers converted the characteristic key figures of different systems, such as mass and energy requirements, to a common denominator: one kilogram of human muscle can generate around 50 watts of power, while a technical actuator of the same weight, such as a hydraulic cylinder, can easily generate ten times that amount. In fact, when analysed, the individual robotic components often proved to be superior to the biological ones - drives perform better, sensors are more sensitive and support structures are more robust. However, if you have to assemble them into a complete system, supply them with energy and possibly also with hydraulic fluid and cool them, the calculation looks less favourable.
The 22 robots capable of walking at a typical human pace did not perform well. As soon as the researchers put speed in relation to mass, size or endurance, the bipeds, which have been optimised by evolution for walking, were able to hold their own against the machines. Individual robots specially designed for this purpose still managed a respectable performance. One specimen in particular, called Cassie, which was developed at Oregon State University, could outrun humans - although the robot is practically only made up of two legs.
One of the best humanoid robots currently available with legs, torso and arms is "Atlas" from Boston Dynamics. While Cassie needs around 200 watts to walk (humans need around 450 watts), Atlas consumes ten times as much at the same speed. In a race against humans, Atlas would be outclassed: He can't run as fast, nor can his battery last longer than an hour.
Evolutionary optimised for efficiency
As one possible reason for the discrepancy, Riener's researchers cite the fact that biological tissues contain elastic structures. Connective tissue such as tendons and ligaments absorb shocks, store energy and can later release it again in a useful way. According to the scientists, most humanoid robots do not contain any such components - at least not yet.
The scientists came to similar conclusions when they looked at climbing stairs, lifting or gripping objects or other special movement sequences. There are always some robots that master individual activities, but none of them are as versatile as humans.
However, the authors note that robots have become increasingly skilful and powerful in recent years and expect further breakthroughs. Furthermore, it is not necessarily desirable to design a robot that masters as many disciplines as possible. Analogous to the division of labour in the human social structure, specialised machines could also master certain tasks together as required. Nevertheless, even for such scenarios, it would still be necessary to initially accommodate more functions in the individual robots - otherwise the networks would become too large and heavy.
With new concepts for more compact designs
Riener and his colleagues have not yet found a clear answer to the question of why it has not yet been possible to construct a superior robot from the faster and stronger technical components. "One limiting factor," they speculate, "could be bringing the structural components to human scales and integrating them into a compact design." Our muscles already contain the energy supply in the form of molecular energy carriers. Artificial actuators need bulky batteries, cables, compressors and hoses. There are also lots of sensors where we have fine nerve tissue. Nevertheless, machine learning could help to make movements smoother, more predictive and more efficient. However, this will only make a robot less clunky to a limited extent.
The future of robotics lies in cooperation, not competition
A way to achieve significantly more compact integration, according to the researchers, could be to rely on fundamentally new principles in the designs in addition to currently already highly developed components such as electric motors and hydraulics. For example, the field of soft robotics has been making promising progress for over a decade. Nevertheless, it would not only be a challenge to bring such concepts to application maturity, but also to connect them to the existing extremely powerful components using suitable interfaces.
Thus, despite some impressive and sometimes frightening promotional videos from robotics companies, there is still no overpowering Terminator in sight. No machine can replace humans when it comes to a wide range of challenges. On the contrary, the developments are even encouraging: robots are now becoming just autonomous enough to support humans in everyday tasks that would be too strenuous, dangerous or small-scale for us without the need for overly complex controls. The future of robotics lies in cooperation, not competition.
Spectrum of science
We are a partner of Spektrum der Wissenschaft and want to make well-founded information more accessible to you. Follow Spektrum der Wissenschaft if you like the articles.
[[small:]]
Cover image: © AndreyPopov / Getty Images / iStock (detail) A robot would probably win the tug-of-war - but only if it was built precisely for this purpose.
Experts from science and research report on the latest findings in their fields – competent, authentic and comprehensible.