Artificial anatomy.

What happens if a robot passes the Turing test (the test of a robot's ability to replicate intelligent human behavior [Turing, 1950])? Which behaviors or actions would they be capable of performing with each other and/or with human beings? How would knowing that a percentage of those with whom we relate are not entirely flesh and blood modify our own conduct and behavior? Director Ridley Scott's seminal and excellent film “Blade Runner” (Warner Bros, 1982) introduced a whole generation to the details of how to appear human while being a robot. In the film, the bladerunners had to detect, capture, or annihilate robots whose bodies and minds were virtually indistinguishable from our own. To do so, they put subjects through a test that essentially recapitulated the Turing test. It consisted of asking questions whose answers would allow us to distinguish whether the answerer is human or robot. The screenplay, based on the Philip K. Dick's short tale, “Do androids dream of electric sheep?” (Dick, 1968), puts in the hands of the main character, Rick Deckard, the responsibility that the most advanced versions of androids not escape search and capture, and therefore, not pass the Turing test. Somewhere in the background is the concept that there will be at least some human emotions or behaviors that a robot will never be able to fake well enough to fool a human. Although the above questions are very relevant, with a huge and constantly evolving philosophical component, they are difficult to tackle. There are, however, two aspects that can go unnoticed in this field. The first is the most difficult: the meta-problem of how an android would “feel” this problem. Although that problem is out of the scope of this volume, we will deal with a variant of it in the next few paragraphs. The second aspect is already contained in Dick's own original story and is masterfully represented in Blade Runner. In both works, it is assumed that artificial bodies will be totally indistinguishable from human bodies, both in form and in physiology. Artificial bodies could even be better, without it being an apparent problem for humans or, of course, for robots. Thus, while artificial intelligence (AI) advances month after month toward an unstoppable universe in which programs will reach ever closer to the frontier of passing the Turing test, it seems that the frontier of ensuring that bodies can be equipped with artificial anatomical elements that mimic human anatomy and physiology proceeds very differently, both in speed and conceptually. The scientific fields to which we refer are neurorobotics and neuroprosthetics. We understand neurorobotics as the set of artificial systems designed to imitate the computational models of living neural networks, including their algorithms. Neuroprosthetics is the engineering discipline devoted to making possible the relationship between such algorithms and the physical devices that will execute the movements. Thus, the operability arising from the combination of artificial movement systems similar to humans, with the algorithms and neural network models that control them, primarily seeks to generate (increase or replace) the movement functionality of human beings. Such a frontier advances more slowly than AI, on the one hand, because a qualitatively and quantitatively different investment is being allocated in each race. On the other hand, there is no doubt that the question of how the subject “feels” the new functionality is now a not insignificant but philosophically irrelevant problem, because the subject is “mentally” the same human being (and not an android foreign to our humanity). Still more relevant is the fact that the artificial anatomy and functionality of the device do NOT currently need to be indistinguishable from humans. There is nothing wrong with perceiving, both the subject and the environment around him/her, that someone has a bionic hand, arm, or leg. Or that one walks (or even runs) thanks to an exoskeleton that is beautiful or cumbersome to varying degrees (read, for example, Tan et al., 2021). The application of neurorobotics and neuroprosthetics after diseases or accidents will allow future restoration of brain function and limbs that today have no repair or treatment (see for example Masteller et al., 2021; Micera et al., 2020, or Kansaku, 2021). Two of the most challenging problems of neurorobotics and neuroprosthetics today are: (1) knowing enough about what we humans do to be able to imitate it; (2) combining what we know about human anatomy and physiology with the design of a prosthetic system that mimics it. This AR Thematic Papers (ThPI) Issue presents a representative sample of current efforts. For this, our guest editors (GEs) Filipe Oliveira Barroso (Juan de la Cierva postdoctoral researcher), Diego Torricelli, and Juan C. Moreno (both Científico Titular), have compiled five insightful papers in these fields. The three GEs work at the Neurorehabilitation Group of the Cajal Institute (CSIC) at Madrid, Spain. The works by Radeleczi et al. (2022), Martín-Caro Álvarez et al. (2022), and Okajima et al. (2022) analyze knowledge about what humans do with our arms and legs, always with an eye toward learning the most relevant aspects to be mimicked in artificial devices. The work by Figueiredo et al. (2022) is a perfect example of how an exoskeleton is fed back with the user's experience and changing external circumstances to produce a more adequate performance of the exoskeleton, and finally, in Demofonti et al. (2022) we assist the analysis of how our knowledge of the anatomy and physiology of the human upper limb is combined with the design of a prosthetic system that imitates it. A common principle underlying all these works is to find out what the human body does to imitate it, as well as to combine our knowledge to design prosthetic anatomy with a neurorobotic algorithm capable of interacting with the environment and producing the same control of subject-environment interaction as in humans. As our GEs mention (Barroso et al., 2022), the challenge is to generate a device capable of detecting errors made and to be able not only to generate a movement in accordance with the requirements of the situation, but also to minimize the potential error in the resulting movement of the device. This is exactly what the human cerebellum does naturally (Peterburs & Desmond, 2016). Unlike the android's minds in Blade Runner, in this film, an additional parallel micro-story is told: how an expert in bioengineering makes artificial eyes. In his case, no one cares if the eye appears artificial, but rather whether the eye can see. And in his case, the eye resembles the human eye not because of aesthetics, but because in no case is it more obvious that the physiology (of the human eye) derives from the anatomy. As Antonio Machado, the great Spanish 20th-century poet said, “the eye is not an eye because you see it, but because it sees you” (Machado, 1924). And with a hefty dose of cinematic license, let us say that Deckard might well subscribe to the idea “what matters is whether the exoskeleton walks, not whether it looks like a human leg”, though for it to walk, it probably would have to resemble a real leg. If it walks, it does not matter if it passes a Turing-like test or not. José L. Trejo: Conceptualization; writing – original draft.