July 8, 2020

Artificial intelligence: The Navarrese physicist who teaches machines to read your mind | Innovation

Humberto Bustince (1958, Ujué, Navarra) ended up dedicating himself to the artificial intelligence (AI) Natural way. Although he studied theoretical physics and the power always attracted him quantum mechanics, due to life's circumstances, he ended up doing his doctoral thesis on intuitionist logic to represent human reasoning. “I was so passionate about the subject that I began to be interested in taking it to AI,” he acknowledges now, many years later, at the gates of receiving one of the research awards delivered by the Scientific Information Society of Spain in collaboration with the BBVA Foundation.

With this idea in mind, he decided to enter a field that then had little travel: approximate reasoning. “Computational intelligence is a subfield within AI that consists of three pillars: how machines learn through neural networks, how to develop evolutionary and biological algorithms so that they can give answers in real time and how we relate to them, an area associated with the natural language ”, delimits. "This last point is where the approximate reasoning comes in."

This branch of AI, in which intelligent text and voice processors are framed, is based on the ability to draw conclusions from incomplete information or with some degree of uncertainty. Bustince quotes the former president of the CSIC Enrique Trillas to remember that the human being tends to function by approximate reasoning and fuzzy logic and, for most of the day, it is not expressed exactly.

"In classical logic, things are true or false, but, in real life, not everything is black or white," he says. “People don't say it's 12 or 27 degrees, but it's cold or hot. In the jungle, nobody would say with certainty that when you are more than 100 meters from a lion there is no danger; we are seeing, as we approach the lion, if the distance is or is not dangerous ”.

To reach these conclusions, it is essential to understand the context. It is not the same to deal with problems using weak AI or strong AI. Bustince explains that the first frames the algorithms that enable autonomous cars and those who are giving very promising results in oncology. The second goes a step further and aims to decipher feelings, sensations, emotions and perceptions.

People do not say that it is a temperature of 12 or 27 degrees, but that it is cold or hot.

"To be able to advance in this field, biologists and doctors are being involved that help to understand what part of the brain makes one or the other decisions and how the chemical reactions are," he explains. "It is an area that is still being born, but it is moving fast."

If we read a story about Donald Trump in which he is not mentioned directly, our memory is organized in a way that allows us to understand that we are talking about the president of the United States, who is a real estate mogul who has been in power for about three years. The idea is that the machine is also able to reach these conclusions. This is where data fusion comes into play.

"When we analyze a student's grades, we can intuit that If you have good grades in math, you may also be good at physics, although it will not necessarily be good in history, ”Bustince explains. "For a computer system to establish these relationships, it must analyze both the similarities that exist in the data and their differences."

Four years ago, the fusion of data led this physicist to start working in one of the areas with the greatest projection of AI: the computational brain. The research team that coordinates at the Public University of Navarra (UPNA), where he works as a professor, together with researchers from Australia and Taiwan, he developed a system that monitors brain activity and translates certain signals emitted by the brain so that they can be interpreted by a machine.

To train him, they asked a large group of people to think about moving their left or right hand and monitored their brain activity. Now, your system guesses which hand someone is moving with 95% accuracy thanks to the fusion of data from different wavelength bands. And it doesn't just work with binary options – left or right, yes or no; Your system can work with up to four different alternatives.

“This type of research gives rise to interesting applications in medicine, for people who are aware but cannot communicate. From this project, we have developed tools so that these people can tell you what they prefer to eat without having to move or open their mouths, ”explains Bustince, who also confesses that, among those interested in this development, there are telephone companies that are looking for , in a decade, your mobiles can be turned on and off with your mind.

. (tagsToTranslate) physical (t) navarro (t) teach (t) machine (t) read (t) mind (t) researcher (t) humberto bustince (t) apply (t) reasoning (t) approximate (t) intelligence (t) artificial (t) functioning (t) be (t) seem (t) brain

Source link