Carme Torras is a research professor at the Robotics Institute, where she leads research into perception and manipulation. As well as her scientific career, Torras has written a unique work of literature. She is the author of La mutació sentimental, which won the Premi de Ciència-Ficció Manuel de Pedrolo 2007.
© Pere Virgili
You are one of the best-known researchers in robotics in Europe, a relatively new field. How did you get into this discipline?
Once I finished my degree I started working in a multinational, but I was more motivated by research and wanted to keep learning. Thanks to a Fulbright scholarship I was able to study at the University of Massachusetts, Amherst. At that time what interested me most was the brain, artificial intelligence. At Amherst I focused on brain modelling, and Professor Michael Arbib was my supervisor for my Master’s thesis.
And what was the focus of that thesis?
I studied neuronal modelling. We worked with neurologists from the Ramon y Cajal Hospital in Madrid. Specifically, I modelled the nervous system of a crayfish. With very few neurones you can model learning, on both the physical and chemical level. The aim of my research then was to understand how we learn, how the brain acquires knowledge.
It cannot have been easy to follow this line of research.
It was difficult, because it needed a lab that we did not have. I started to look at artificial intelligence and robotics. At that time the Robotics Institute, which had been founded by Gabriel Ferraté, was known as the Cybernetics Institute, and was a joint centre between the Spanish National Research Council (CSIC) and Polytechnic University of Catalonia (UPC), the two institutions for which I work. Currently the institute has four fields of research. First, there is robot kinematics and design, the closest to mechanics. Second there are mobile robots, focusing on urban outdoor robots. Then there is process control, which deals with water distribution, electricity grids, fuel cells, etc. And finally, there is perception and manipulation, the group which I direct. Initially I worked on the kinematics and programming of industrial robots. Now, instead, I look at cognitive robotics with social applications.
What does cognitive robotics consist of exactly?
In the application of artificial intelligence techniques to robotics. And it sets out to mechanise in a robot actions that humans do naturally, like task scheduling or the perception and manipulation of objects.
What do you understand by perception and manipulation?
We mean that the robot has to be able to make itself a good representation of the environment and the user. We work in a domestic environment, in robotics for care and services. This has very different determining factors from industrial robotics. First, the robots that move in human environments must be safer; they cannot make sudden movements that could pose a danger to the user. Secondly, a non-expert must be able to programme them. For example, the person can teach the robot how to beat an egg with a simple demonstration, and the robot must be capable, thanks to cameras inside it, of acquiring this information and learning the basic skills necessary.
Have you created a robot that can cook our dinner?
The cooking robot is still at the research stage. Not long ago we carried out a European project, called Paco-Plus, where we managed to get a robot to clear the table and put the plates in the dishwasher. It was not intended to be transferred to the private sector, but it involved up to nine research centres.
Was it an international project?
Yes, we are completely integrated into the international robotics community. Recently, we have participated in three European projects in the field of cognitive systems. The Paco-Plus project, directed by the University of Karlsruhe, involved groups from Denmark, Sweden, the Netherlands, Belgium, the UK and Slovenia. The GARNICS project, also headed up by a German group, was aimed at creating a robot gardener capable of taking samples in big plantations to optimise the yield. And the goal of the current IntellAct project, an acronym for Intelligent Action, is to teach robots to carry out simple manipulation tasks simply by doing them in front of them and correcting them when the reproduction by imitation is not quite right, just as you would with someone who was learning.
Could domestic robots have feelings?
There are people who work on giving cognitive robots emotions, to make them more sociable and friendly. For example, in Japan they are already working on various robot carers that can look after elderly people or children, something that I question. It is good that robots can expand our capabilities, but I do not like the idea that they are used as baby-sitters. In the long term it could be harmful if children have robots as slaves, and are not obliged to negotiate as they would when playing with other children. How will they learn empathy if they do not have someone in front of them who responds emotionally?
What kind of robots should children have, then?
A robot, to a greater or lesser degree, can affect the development of a child’s interpersonal relationships. A child who only plays with a robot will not have the emotional feedback needed to generate empathy. We learn emotions by seeing them in others. In the end, as Robert C. Solomon says, the relationships that we build are those that in turn shape us.
But it is also true that robots or games can stimulate multitasking [the ability to execute different tasks simultaneously], an ability that is well-developed in the younger generations, but which is detrimental to the ability to concentrate, because they are ready for a thousand stimuli at the same time. It is a vital attitude that trains the sensory motor skills, the driving reflexes, stimulus response, but which decreases the capacity to concentrate deeply on a problem. You see this, too, in the new generations of researchers: the machines generate such precise graphs and the algorithms give such automatic responses that often the researchers lose the physical sense of what they are doing.
Today people are still more dangerous than robots.
The danger is not that the robots will become more human and attack us, but that humans become robots, that they limit their actions to the simulated world inhabited by robots. Robots should increase the abilities of people and give us more autonomy, instead of decreasing it.
In your book La mutació sentimental, set at the beginning of the 22nd century, there is a robot that has a creativity prosthesis, designed to challenge the user and stimulate their wits.
What we must ask ourselves is whether we want robots to do the work and sideline humans or, conversely, if we want robots that will stimulate us and make us grow as people. “Spoiled robots make spoiled people, robots slaves make tyrants and robot entertainers remove the brains of their owners,” says a character in the novel. We should have an opinion as to what kind of robots we want, otherwise we will have no control over what we get sold. We have to start to think differently about the use and the point of robots.
Machines are changing our cognitive capacities and it is important to be aware of this to decide which capabilities we want to have. We cannot easily control our reaction to stimuli, but we can pick and choose which stimuli we want to receive and for which we want robots to be designed. And in this, society has a lot to say, because companies will sell what they want to, and the users must know how to discern what really suits them, whether we are talking about a doll or a butler.
A lot of education will be necessary to develop this awareness.
Yes, I started to write La mutació sentimental exactly for that reason. There are scientists who still aspire to creating a fully autonomous robot, not dependent on humans and capable of organising itself and setting its own goals; they even make robots with bodies programmed to keep growing. I see no sense in encouraging this kind of experiment without asking ourselves what will happen afterwards. In other words, there are those who think we will not have a fully developed artificial intelligence until we have created completely autonomous systems. I, on the other hand, think that robots should serve people. For me it makes more sense, for example, to design robot doctors who are capable of carrying out surgical procedures with more precision than us, than inventing a robot that has its own goals.
All this raises fascinating ethical questions. Just as overcoming the limits of biology has given rise to bioethics, in your field there is now roboethics…
Yes, more and more researchers are becoming interested in the subject, and the potential users are also paying more attention to it. At the previous robotics world congress there was a session open to the general public, dedicated to human–robot confluence. It was a very enriching experience contrasting very varied viewpoints.
What limits can we impose on the entertainment industry in the development of robots and humanoids?
I often use the example of Tamagotchis, devices which for a time were all the rage among kids. They were like living things, which needed feeding every day so they did not die. Tamagotchis were a hit and sold loads, but for me they are a clear example of an artefact that gives us nothing and creates a useless dependence. On the other hand, there are educational robots that many schools have integrated into the curriculum. At the moment, they are holding the First Lego League, for example, a worldwide competition for children aged between six and nine and ten and sixteen. It is a team competition. Some do the design, others programme and some take care of the sensory aspect. With these robots we can provide examples for explaining physics or mathematics. Last year three Catalan girls trained in our workshops won an international prize.
Who formulates the discourses that feed roboethics? It is an open debate or restricted to engineers?
On the contrary, it is more important now than ever to open the debate to other disciplines. Like I said before, at the previous robotics world congress, in Karlsruhe, there was a forum entitled Robotics Meets the Humanities. At the table were two robotics professors, two from humanities, a philosopher, a science fiction filmmaker, there was Marcel·lí Antúnez as a performing artist who uses robotic technology, and I was the moderator. The robotics community has already glimpsed the need to draw up possible future scenarios, which give us ideas for research and enable us to assess whether they are feasible and where they will take us on a human level.
Do you think autonomous systems should have rights…?
There is not much sense in conceding rights if they have to serve people. Instead they should have obligations. What does need legislating is, if a robot harms someone, who is responsible? The engineer who designed it? The company that sold it? The person who was using it? Think about drones, military robots.
Can we talk about military roboethics?
A friend from NASA explained to me that there are already veterans from Afghanistan and Iraq who fought using remote control drones. In combat their emotional involvement was not as intense as if they had been on the battlefield, but they suffer anxiety attacks and end up having psychological problems as serious as those of Vietnam veterans. If a military offensive is trusted to robots, knowing that they may impose limits, it would probably limit the scope of the destruction, provided that both sides rely on robots and they attack each other, of course. There is much to talk about here…
What role will robotics have in the smart cities of the future?
Robots will be useful for doing all kinds of tasks. We have sensors in waste containers that tell us when they are full, for example. Rubbish collection will be robotised. Lots of other infrastructures will also work autonomously. There is a project called Roboearth that consists of making an internet for robots, so that each of them can add their expertise and knowledge into a central system, a data cloud. Then a robot can go somewhere and know exactly where all the devices it has to use are, even the first time it sets foot there, because it will have access to all the information that has been modelled. We are also in contact with Barcelona City Lab. Here we have The Humanoid Lab, a workshop for university students to start programming robots.
How do you see Barcelona in the field of smart cities? Do you think it is high in the rankings?
Yes, very. Our institute took part in URUS, a pioneering project on the development of robots that guide users around urban environments. It was undertaken in parallel with a similar study at the University of Osaka, and the two cities were twinned as promoters of urban robots. There was a demonstration in Passeig de Sant Joan: two mobile robots and an automatic car. People in the street approached and asked questions. You could ask the robots where the hospital was or ask them to accompany you to the automatic car, which would then drive you to your requested destination.
Barcelona has signed the Smart City Protocol and the City Council is working to consolidate good practices to make Barcelona a smart city. What is the view of the Robotics Institute on this, and how is it participating?
Well, in fact, URUS drew up a preliminary protocol to define everything that had to be standardised when defining actions and processes in different places. If people had to be evacuated in the middle of a fire, for example, and we had robots that helped the humans in this emergency, it would be good to have homologous behaviour everywhere, both for reasons of efficiency and so that their actions would be understandable to all of the users. If we standardise the protocols, we could also make the robots more universal, and therefore make the investment needed for developing them more cost effective.