Antonio Torralba, MIT researcher: “We don’t create intelligent machines to be independent, but to improve ourselves” | Technology

Antonio Torralba (Madrid, 1971) believes that “the human brain is the most interesting and complex thing in the universe and that artificial intelligence (AI) tries to understand how natural intelligence works and how we ourselves work”. He is considered a pioneer and a world reference in computer vision research and in the application of AI in this field. One of the most important lines in his work has been to understand objects in the context of the scene where they are found and in relation to other objects. At a time when vision systems are beginning to come quite close to the capabilities of a human being to recognize objects, his next challenge is to study vision not as an isolated sense, but as a whole of all the other senses. He studied Telecommunications Engineering at the Higher Technical School of Barcelona (ETSETB) of the UPC, school in which he was awarded a doctorate honorary cause last Friday on the occasion of the center’s 50th anniversary.

Ask. How would you explain what artificial intelligence is?

Answer. It is still the attempt to artificially reproduce what we associate with natural intelligence, the ability to solve problems and adapt to new situations. In general, it is more difficult to understand how AI works, but although it is technically a sophisticated system, its basic principles are very simple: it learns from the examples of associations that you want it to establish. Currently, they work with neural networks, a simple imitation of the human brain, which consists of a process where the machine decides how the different neurons that form it are going to connect. These are elementary units that produce operations that would be understood in Primary, but, when put together, they are capable of solving complex tasks based on the learning that the system has done.

P. How does it help improve society?

R. Artificial intelligence has many areas. Health is one where it will undoubtedly have a very important and beneficial direct impact on society, from diagnosis to the use of robotics for surgical assistants. In detecting diseases, the human being is still ahead, but the machine is capable of learning and memorizing many more patterns and perceiving more subtle changes than a person. Now we are working on leaving the specialization to start acting in a general way, as an expert. The machine can certainly outperform an untrained person; at the level of the expert, it has not yet arrived, but it is beginning to approach. Another field where it can help is in the prediction of new molecules and the discovery of new antibiotics and treatments. Based on analyzing patterns of known antibiotics and their curative effect, the AI ​​proposes a series of alternatives whose effect is already predicted before testing it. The success rate obtained is high enough to speed up the process.

P. What applications can it have in education?

R. In education, it has two aspects: education in artificial intelligence and as a tool to teach. The first, already very necessary in the present, allows generating common links between subjects without apparent connection. You can connect math with language by turning words into numbers to parse through math operations. In addition, AI can contribute to the educational challenge of adapting to the individual, making it possible to have, apart from the teacher, an artificial teacher optimized according to the abilities of each individual.

P. Do you think that artificial intelligence can surpass human intelligence?

R. Technology has always been an instinct to overcome our own limitations. We have created faster machines, that fly or that are more intelligent in certain problems, but not so that they are independent of us, but to improve ourselves as a society. Deep down, we already live in an environment where there is an intelligence superior to ourselves: society as a whole, and AI is a new element of this community.

P. Can artificial intelligence be dangerous?

R. The great danger of technologies is their harmful side effects, which do not necessarily arise from bad intentions. AI allows for highly personalized attacks, for example, but to prevent them, what is important is good communication between sectors. Politics is necessary to establish the essential regulations so as not to stop innovation and to give companies the tools to avoid possible harmful side effects. In the United States, the leading country, along with China, there are many more resources than here. The list of differences between Spain and the US is endless: there, research has great weight, both at a political level and in the industry, a very important funding engine. There is a lot of interaction with the universities as well, which doesn’t happen much here. The fact that outside of Spain there is access to many more resources causes the so-called brain drain, but the researchers who stay here are as good as those who leave.

You can follow EL PAÍS TECNOLOGÍA at Facebook and Twitter or sign up here to receive our weekly newsletter.