The ancient Greeks used the term Banausia (from banausos, "craftsman," "manual, mechanical work") to refer to manual work and mechanical arts in general, and it had a negative meaning: craftsmen, or whoever performed a manual work, were considered inferior to those engaged in an intellectual work.
Many centuries later, between the 1400s and the early 1700s, the European culture reassessed manual techniques. Some of the procedures used by technicians and craftsmen to modify nature turned out to be useful and beneficial to understand the natural environment. The defence of mechanical arts from the accusations of unworthiness and the refusal to make practical activities coincide with the concept of slavery led to a historical cultural turning point: the end of an elitist image of science and of the distinction between knowledge and skills. In the reassessment of science and mechanical arts, a major and original role was played by Francis Bacon. In fact, he wrote an important and lucid critical treatise on the experimental method and on the good or bad use of science and technology. In Novum Organon, he talked about the condition preliminary to all scientific works: that is, the removal of idóla, namely advance information or prejudices that pollute scientists’ mind and their objectivity. Said idóla were divided into tribus (typical of everybody), specus (typical of the single individual), fori (related to controversies and verbal disputes) and theatri (due to philosophical, religious, cultural dogmatisms). In the same treatise, Bacon stigmatised the existence of two opposite anti-scientific behaviours which he described as being similar to the typical actions of spiders and ants: rationalist dogmatists, lacking contact with reality, are like spiders, that spin webs from themselves; empiricists, lacking theoretical foundations, are like ants, that simply accumulate and use rashly. True scientists combine both theory and experimentation, like bees, that take material from flowers but then have the ability to convert and digest it.
Hannes, a robotics-derived prosthetic hand developed by Rehab Technologies Lab with the collaboration of the Italian Institute of Technology (IIT) and the National Institute for Insurance against Accidents at Work (INAIL)
Mechanical Man, an image digitally created by illustrator Bill McConkey with the photomontage of wind instruments
The I-Cub, an IT-developed robot designed to support research in the field of artificial intelligence
The R1, an IT robot designed to work in domestic and professional situations
jk Illustrations of mechanical prostheses contained in Instrumenta chyrurgiae et icones anathomicae, by Ambroise Paré, 1564
Finally, in De Sapientia Veterum, Bacon brilliantly used the myth of Daedalus to talk about the constituent ambiguity of technology. Daedalus built a device to enable Pasiphae to mate with a bull; this pernicious use of technology gave birth to the Minotaur, devourer of men. At that point, Daedalus made a good use of his intelligence and built a labyrinth in which to confine the Minotaur. The labyrinth was also provided with a safety system, Ariadne’s thread, that allowed Theseus to find his way out. The metaphor is clear: science and technology can be used against or in favour of mankind; therefore, scientists must be responsible and forecast remedies and limitations of the possible negative outcomes of their discoveries. Despite Bacon’s ideas are more than 400 years old, they are extraordinarily topical; in particular, the intuition concerning the ambiguity of technological progress is perfectly apt with the issues concerning the development of robotics and artificial intelligence (AI).
It is not easy to define what a 'robot' is, considering the rapid and continuous development of robotics. The word ‘robot’ was introduced in 1920 by Karel Capek, a writer who went beyond the concept of ‘automaton.’ In fact, he introduced the idea of an artificial machine built by humans to perform precise functions related especially to work (in Czech robota means forced labour).
Over the last sixty years, robotics has progressed extraordinarily. Initially, its products consisted of mechanical, static, passive, repetitive and executive objects; today, robots are becoming autonomous and mobile realities capable of performing not only specific functions, but also general ones. They can be provided with learning and adaptation skills, and act autonomously, without the control of an operator.
The most advanced robots have cognitive abilities similar to those of primates. They are able to communicate through the recognition of words, and can have expressions in their outward appearance that imitate several human emotions.
Currently, there is much debate on the possibility to realise robots provided with an advanced artificial intelligence (AI) such to be able to develop decisional abilities and self-determining processes similar to those of humans. In actual fact, in the collective imagination and in the representation given by the mass media, literature, movies and TV series, robots are increasingly viewed as entities provided with a mechanical body that thinks and behaves like humans. In real life, it is not that simple. Indeed, it is necessary to better explain these concepts.
Mechanical Man, immagine realizzata digitalmente dall’illustratore Bill McConkey con il fotomontaggio di strumenti a fiato
To begin with, it is important to highlight a simple and univocal classification of autonomous and intelligent machines with their different characteristics. Often robots, humanoids and artificial intelligence are considered all alike, but that is not so. First of all, it is necessary to make a distinction between two large types of machines: those provided with a body (embodied) and those without a body (non-embodied). Secondly, it is necessary to verify if machines, both the embodied and the non-embodied, are provided with some form of artificial intelligence (that is, if they are stupid or intelligent).
Embodied machines (provided with a body) are capable of moving and performing physical work. These machines are well-known: scrapers, automation systems and all the technologies that replace humans in physical work, or help humans increase their performances (for example strength, precision, speed of execution, etc).
Usually these machines are “stupid”: they are programmed to work automatically, to perform demanding or repetitive activities with the aim to increase productivity and the performances of the operators using them. Since they are controlled by humans (or programmed by humans), they do not take decisions autonomously: their actions depend on programmes or on human operators. The impact of this robots on workforce, and particularly replacing humans in routine manual works (eg manufacturing) is a very debated issue.
In recent years, some embodied machines have been provided with artificial intelligence acquiring increasing cognitive and decisional abilities. They can be non-anthropomorphic machines (for example self-driving cars), or actual humanoids developed to interact with humans and support them in various environments, such as at work, at home or in hospitals.
The ability to make autonomous decisions, without the control of an operator, is a great technological challenge, which anyway gives rise to many questions from an ethical and regulatory viewpoint. In fact, although this type of intelligent robots is designed to replace humans in dangerous situations, or help them in case of need, it is important not to underestimate the problem of their impact on workforce, not only on routine works. It is also important not to neglect the issue related to the future cohabitation of the two species: humans and robots, both “thinking” but with totally different logics and functioning.
Illustrazioni di protesi meccaniche contenute in Instrumenta chyrurgiae et icones anathomicae, di Ambroise Paré, 1564
Non-embodied technological products are not able to perform work or to make movements, and belong to technologies commonly called digital, ranging from telecommunications to artificial intelligence. Also for this type of products a distinction can be made similar to the previous one. Some non-embodied machines are “stupid,” such as TVs and radios; they have become objects of daily use for years now and humans are almost addicted to them. They process and transmit information, both audio and visual, and have opened the communication sector to modern society. Non-embodied machines have gradually become increasingly “intelligent”: from smartphones to supercomputers, they can perform calculations at an extremely high speed, from several million operations per second in the case of smartphones to quadrillions of operations per second in the case of supercomputers. This has taken place in the last years thanks to an increasing miniaturisation of integrated circuits, allowing electronic devices to perform an increasing number of operations per second, the electric power consumed being equal, and to memorise an increasing amount of data in mass memories.
Over the last 50 years, the progress of electronic technologies has followed Moore’s Law, on the basis of which (about) every 2 years the number of transistors on an integrated circuit doubles, passing from several thousand transistors in 1970 to about 20 billion transistors in 2016. At the same time, the progress of manufacture has allowed to reduce the electric power consumed by transistors proportionally to the reduction of their size (according to the so-called Dennard Scaling rules). This has led to a constant and uninterrupted development of computational abilities in more recent years. The increase in the speed of calculation and in the ability to store data has allowed to realise electronic devices more and more sophisticated, determining the current digital revolution. Internet, research engines, mobile phones, social networks, Big Data, Industry 4.0, digital wholeness, forecasting models in the financial, social, medical, climatic fields are all direct or indirect consequences of the technological evolution which has characterised transistors.
Today computers can process an enormous amount of data, analysing them statistically very quickly and applying mathematical models that allow to forecast future situations and scenarios (in the economic, medical and climatic fields). At the same time, they can also imitate the cognitive processes of the human brain, and create what is commonly called “artificial intelligence.” This is what gave origin to the research engines we all use, as well as to the artificial intelligence capable of beating humans at chess or outdoing them in other activities strongly computational.
However, how can the computational power of a computer be compared with that of the human brain? The performance of computers is measured in FLOPs (Floating point operations per second), that is the amount of operations that a computer can perform in one second. Today’s most powerful supercomputers are able to perform dozens of PetaFLOPs, that is dozens of quadrillions of operations per second.
In 2017 the top 5 computers in the world as to power of calculation were developed in China (Subway TaihuLight with 93 PetaFLOPs, and Tianhe-2 with 33.9 PetaFLOPs), Switzerland (Piz Daint with 19.6 PetaFLOPs), Japan (Gyoukou with 19.1 PetaFLOPs) and in the USA (Titan with 17.6 PetaFLOP). Their consumption of electric power is tremendous, and ranges from China’s TaihuLight with 15.4 MegaWatt to the USA’s Titan with 8.2 MegaWatt. It is interesting to notice that regardless of the records of the single machines, the USA is the country that holds the overall highest power of calculation in the world: currently, 46% of the global power of calculation is American, owing to a wide network of supercomputers disseminated on the national territory. Following there are China and Japan with 8%, and Germany with 7%.
Such a high power of calculation allows to process quadrillions of instructions per second, which is close to the calculation ability necessary to simulate with precision complex biological organisms. We have to remember that it is not possible to measure the power of calculation of the human brain in PetaFlops, essentially because the functioning of the brain is not based on digital electronic operations. However, the instrument of comparison that can be used is an empirical unit of measure called MIPS (Million Instructions per Second), that is the number of instructions per second that can be processed by a processor, be it biological or artificial. A capacity of 1,000 MIPS is sufficient to reproduce the complete functioning of a complex organism such as a lizard, while 1 billion MIPS is the minimum amount necessary to simulate humans. The development of increasingly powerful computers and software is reaching millions of MIPS; we are getting closer and closer to biological performances, but with an incomparably higher energy consumption.
The availability of increasingly powerful calculation machines is constantly extending the limitations of artificial intelligence. At the same time, it is allowing the development of increasingly performing embodied machines (provided with sight, touch and biomechanical abilities), making realistic the assumption that robots are characterised by performances increasingly closer to those of humans.
It is interesting to notice that until the physical and intellectual performances of machines developed separately, the latter did not constitute a source of apprehension for us. We never feared that machines such as computers could be faster than us in performing calculations, or that robots could be stronger, more rapid and more precise than us in carrying out physical work. On the contrary, a large part of our industrial progress has been based on the use of machines that, subdued to our purposes, have increased human performances in specific fields.
R1, robot dell’IIT progettato per lavorare in ambito domestico e professionale
The fact that the physical and intellectual “powers” of machines have always been separate has made us feel safe: computers “think faster than us, but they cannot move,” robots “are stronger than us, but they cannot think;” for a long time, these paradigms have safeguarded the supremacy of humans over their technological creatures. Robotics and artificial intelligence are two worlds created from different realities and technologies (mechatronics and computer science, respectively). Over the years, they have produced “incomplete” technologies compared to humans, in the sense that their increasingly effective emulation concerned only a part of our potentialities: the physical ones (strength, duration, precision) or the cerebral ones (calculation, memory, logical process), and have become of common use without too many problems.
No one felt threatened by a computer capable of winning at chess, because said computer did not have a body and was not able to do anything else; no one feared a machine capable of raising tons with disarming easiness and precision, because it did not have cognitive abilities. The mutual contamination of the two technologies gave origin to the other species: intelligent machines capable of moving, acting and taking decisions autonomously, making us feel threatened as dominant species. In the collective imagination, if strong robots are also capable of thinking and computers that beat us at chess are also capable of running, humans are in danger because they can lose control over their artificial creatures.
Intelligent and autonomous machines (A/IS Autonomous Intelligent Systems) represent a true technical, scientific and cultural revolution, and are perhaps the greatest consequence of nanotechnologies; they are seriously starting to measure up with our culture, customs and society, while raising doubts, anxieties and fears. We need to ask ourselves a question, though: are these fears grounded? According to our opinion, similarly to all new and unusual realities, this revolution-evolution is not to be feared, rather it is to be studied and understood.