The trajectory of Pau Rodríguez (Sabadell, 30 years old) can only be understood from the forced professional nomadism and academic excellence that has characterized many researchers. A destiny shared with many of his colleagues who, in his case, has taken him from his native Catalonia to Belgium, to study a master’s degree; and right now to Montreal, in Canada, where they have offered him a stay to investigate. On the other side of the Atlantic is where this doctor in computer engineering has seen how his research on artificial intelligence and neural networks gained relevance. A work that has been recognized this week with one of the Young Computer Science Awards of the V edition of the Research Awards granted by the Scientific Information Technology Society of Spain (SCIE) and the BBVA Foundation.
As he explains by video call, he has achieved the award thanks to the fact that he has designed machines that detect human pain with great precision. Artificial intelligence, once it learns from the data entered by doctors and the images it receives, associates certain movements or facial expressions with different levels of suffering. Small grimaces, blinks or hand gestures, imperceptible to the human being and not to an algorithm, which facilitate medical treatments, such as in pain units, and hospital stays. “It is a system that learns through video images. It is more complicated than with photographs because here there are three dimensions instead of two. Time has to understand it too. It took us a year to develop the project, ”says Rodríguez.
But artificial intelligence is only part of the solution. Neural networks work the magic that turns machines into artificial eyes capable of deciphering pain. This deep learning system, which is still a mathematical method inspired by replicating the functioning of the human brain, associates data, such as the content of images, and establishes connections between two neurons. When the conclusion is correct — for example, there is pain — the system is reinforced. If he is wrong, the stimulus would be the opposite, a negative one. “The recent revolution in this field makes it possible to create very deep networks, with many layers. Each layer receives the count of the previous one. The applications are going to be incredible, ”says Rodríguez.
One of the difficulties he encountered was getting the algorithm to attend to all patients, without previously knowing anything. Machine learning requires infinite amounts of data to function optimally. The problem here was that he was studying specific people. He only learned from these models. The formula to solve it, according to Rodríguez argues, was to evaluate the training with a group of patients never seen by the machine. “If it worked well, it was possible that I would learn some general rules that apply to anyone, not just one person. And it worked, ”he says.
A hospital in Nice has incorporated the project as part of its patient care services. They control the pain they suffer. However, they have seen that they can get more out of it. The algorithm helps detect dementia-related diseases, such as Alzheimer’s. Because the training is carried out with video images, it is also easy to observe those movements of the facial muscles that give clues about this type of pathology. “The intention is that technology identifies disease before humans. When a person tries to reproduce a sign or gesture and does it wrong, which is one of the symptoms of dementia, the machine quickly identifies it, ”says Rodríguez.
The ethical framework
With so much artificial intelligence, ethics could not tiptoe. The technical community is increasingly on top of biases, discrimination and misuse present in this type of innovation. Although they are still in force, the zeal has increased when it comes to setting certain limits. Rodríguez himself says that in Canada they take this matter very seriously. In their academic works they force them to fill in a specific point about what the implications of the project are, who benefits and harms, and how far their development can go. “The ideal would be for countries to legislate in this regard, but at least the debate is not left aside and there is a certain will that companies do not create racially and socially biased algorithms,” he points out.
Different studies show that this almost new decade will be dominated by artificial intelligence. Rodríguez is clear that we are in the middle of a revolution, capable of penetrating all of society and in areas, a priori, away from technological boasts. “You will be able to solve many complex tasks. It is already used, for example, in the design of new drugs. It has created molecules that interact with humans during illness and reduce some of the side effects, “he adds.
Rodríguez does not know how far the algorithm can be developed. His intention is for anyone to get the most out of it. For this reason, you have chosen to leave it as open source –open source, in English-. Although I may leave this research a bit out of the way right now, your next project will focus on artificial intelligence requiring little training and no supervision during learning. That is, it replicates brain function as much as possible. “I see him as a child, who directly learns by playing, by trial and error without anyone controlling him,” he concludes.
Awarded by the Spanish Scientific Information Society (SCIE) and the BBVA Foundation
In this V edition of the Research Awards granted by the Spanish Scientific Information Society (SCIE) and the BBVA Foundation, the winners in the Young Computer Researchers category, endowed with 5,000 euros and which recognize creativity, originality and excellence, are the following :
Mikel Artetxe, Research Scientist at Facebook AI Research.
Nicolas Calvo, postdoctoral researcher in the Computer Science department of the University of Almería.
Llogari Houses, Associate Professor at Edinburgh Napier University and CEO of 3FINERY LTD.
Alberto Garcia, Software Engineer (ML&CV Applied Scientist) at Facebook Reality Labs.
Sara Nieves Matheu, postdoctoral researcher at the University of Murcia.
Pau Rodriguez, associate professor at the Autonomous University of Barcelona and researcher at Element AI.
The winners of the National Computer Science Awards, which recognize researchers, public and private entities that have dedicated their work to the study, strengthening and dissemination of information technology, in this edition are the following:
Antonio Bahamonde, Professor of Computer Science and Artificial Intelligence at the University of Oviedo, José García Santesmases Award.
Eneko Agirre, Professor of Computer Languages and Systems at the University of the Basque Country (EPV / EHU), Aritmel Prize.