Ph.D. in Physics from the UAB, Alejandro Turpinexpert in computer vision e artificial intelligence (IA), has spent more than five years training and researching in Scotland, Germany and Australia. Now with a project photonic technologies to implement ‘machine learning’ algorithms and AI, and thanks to a beca Junior Leader of the ‘la Caixa’ Foundation to pursue a postdoctoral degree at the Institute of Photonic Sciencesgo back home.
-Welcome. What does this scholarship mean to you?
-From the personal point of view, stability. Sometimes the world forgets that scientists are people. And on a professional level, the scholarship allows me to continue in a field that is very elitist without having to worry about anything other than developing my ideas.
-‘LAIght: artificial intelligence at the speed of light’. That is his project. How is it born?
-AI algorithms need training that can last for months. The energy, or the CO2 equivalent of the machines that are connected to train them, is much higher than that emitted by several cars throughout their entire lives. We consume a lot of energy. It’s something intangible, but it happens.
And that’s where your idea comes into play.
– There are many of us who are fighting to develop an alternative. Remember that algorithms are based on mathematical operations. What it is about is that these operations are carried out automatically with light. We seek to design complex materials that allow us to carry out such mathematical operations in a natural way directly in the hardware. The idea is that these materials are passive and not connected to any electrical device. Thus, when light passes through a cascade of these materials, it will refract, diffract and transform in such a way that it will implement a series of mathematical operations equivalent to those carried out in these AI algorithms. It is turning everything that had been done up to now. Light interacts with any medium, so we are going to exploit this behavior to our advantage. This is the general idea and concern we had when designing the project. In the background there are other very important problems, and that particularly concern me.
-Privacy. Who else who least has ever used FaceApp. What the app does is record your face and then it modifies it with algorithms. But it has recorded your face, it already has your information. This information always has to be stored on some device, and from that moment our privacy is compromised. Someone could access that data if they ‘hack’ the system. The fact of operating naturally with light avoids this, because we don’t have to store any information, what we store is the final result. The development of these technologies can allow us much greater privacy. It has many promising features to make it very efficient and we don’t have to store information because everything will be processed in real time.
-Would it be a faster, more sustainable and safer AI?
-That would be the Holy Grail of the project. Whether we succeed or not remains to be seen.
Related news
-For when do you calculate to be able to have any results?
-We are having them, it is not a Eureka! They always take very small steps. Deeptech projects have a fairly long development timeline and need a lot of funding, but they can have a very good outcome, they can generate a lot of know-how and many companies, which is one of the objectives of the Institut de Ciències Fotòniques , that the investigation does not remain only in articles. We want to contribute to society and generate companies that improve the world. For example, one of the objectives that I seek is to develop a series of layers, of physical elements, that when placed in a microscope objective can detect whether a cell is cancerous or not; I want to identify if a person has just fallen to the ground… they are very specific applications. With these projects we are not going to replace the AI that is done in computers, we can complement it and make part of its processing more efficient thanks to these devices.
A future of opportunities
EL PERIÓDICO and the ‘la Caixa’ Foundation give voice every week to the social, cultural and scientific profiles that with their efforts are creating a society with more opportunities for all.