Cristian Canton (Terrassa, Barcelona, 45 years old) was appointed last Friday associated director of Barcelona Supercomputing Center-National Supercomputing Center (BSC-CNS). It is organically below the director, Mateo Valero (to the helm since 2004), but it will be the one to take the day to day of this public institution, a reference in Spain and Europe. In the dear -owned chapel of the center is located, the eighth most powerful supercomputer in the world, according to him. Since February of this year, the BSC-CNS also has developed with 100% European technology.
All this calculation power is dedicated to scientific research of general interest. “We have about 1,000 researchers who work in Life Sciences, Earth Sciences, Computer Science and other topics,” Canton explains. “Here we are not interested in raising the actions of a company, but cureing cancer or avoiding climate change. One of our projects, for example, is dedicated to predictions on pollution flows in cities like Barcelona.”
Doctor of computer vision for the University University of Catalonia (UPC) and an expert in automatic learning and security of artificial intelligence (AI) systems, Canton has worked for the last eight years in the finish line, where he was director of, and previously spent a five years in Microsoft. He will replace Josep M. Martorell, who held that position since 2016.
Ask. It goes from working in one of the main technological to a comparatively tiny and public financing center. How easy is to make the jump?
Answer. I have focused on my career for good. I have never deviated from that. In the finish line, I carried all the part of child pornography, terrorism, organ traffic, drug trafficking, etc. I used AI to detect those contents. Time passed and they made me head of the responsible for the corporation, my goal was to ask me how we make our algorithms fair, auditable, robust, transparent. Therefore, gravitating towards a center like the BSC-CNS for me has been very natural. I will continue trying to help the world, and I will also do it in a place where I will not have the pressure to satisfy the shareholders. The way of working is very different in a public institution. Now I have to adapt.
P. What can be the contribution of the BSC-CNS to the AI, taking into account that the main advances of this technology are leaving the laboratories of the great technological ones?
R. Large corporations have infinite resources, but apply them to commercial objectives. Here we do not have the pressure of obtaining income, so we can focus on issues such as AI applied to study climate change or to apply great language models to genomics. There are battles that we want to win and others that do not. The big corporations will be punched in many things. Fantastic, we need them, especially if they develop open source models, which allows us to learn directly. What we are seeing in recent years is that the generative AI is everywhere, it is already a key technology for the progress of almost all scientific disciplines. Until now, the center of the BSC-CNS was the machine, the Marenostrum 5. Now, with the arrival of the AI, this is the protagonist. We could say that the BSC-CNS is a center of AI that has a large machine that supports it. We have leading researchers in various disciplines and one of Europe’s largest supercomputers, all under the same roof.
P. What is your idea for the BSC-CNS?
R. We have a fragmentation of efforts, people who are working in AI in different fields. We want to see what the common denominator is. We have created, to carry this transformation. The idea is to create a horizontal layer to all these branches of knowledge that causes all ships to rise through the tide, that everyone benefits from AI.
P. The BSC-CNS has been selected by the EU as. What supposes that in practical terms?
R. I think it will be very important for Spain and for Barcelona. Until now, the BSC-CNS served scientists, people who did research, doctoral theses, and so on. Now it will also support small and medium enterprises that do not know how to take advantage of AI to improve their products or services. Startups can train their models here. Maybe they don’t need to buy space in a data center, because they can come here and have it in a month. The idea of AI factories is to accelerate the transfer of AI to small and medium enterprises, as well as administration.
P. Since Trump returned to the White House, in the EU there is spoken of all two things: of rearme and technological autonomy. To what extent do you think Europe manages to be really autonomous from the US or China in technological matters? What role can BSC-CNS play in this?
R. The current geopolitical scenario is very complicated. In addition to Trump in the US, China is there and Asia has the monopoly of the production of the GPUS, the chips that make the AI work. Not thinking about technological sovereignty would put us in a position of weakness. It is important to have at least one strategy and implement it, because if one day they cut your tap, for whatever reason, we would not have the ability to generate. We would be left behind very fast. We are working on chips design. I think it is important to seek sovereignty, both hardware and software, as a mechanism, I would not say defense, but of preparation. In complicated moments, having some independence guarantees that your progress will not stop. The BSC-CNS historically has a great talent in systems design. It can contribute and has contributed to pursue technological sovereignty, for example with startups out of here as Open Chip. It is critical to have the data on European land, managed by us. And have their own foundational models, that represent the values of the Union, built with data that have the correct accreditations, that represent all the languages in a reliable way, etc. We have the talent to do it. I have spent many years in Silicon Valley and where I am always heard conversations in Spanish and Catalan. We also have resources and machines. It is necessary to cook everything well. I want to move the ladle and see what comes out.
P. That the EU is minimally independent in terms of software and hardware would cost a lot of money, much more than is being invested today.
R. Don’t know. Look at. It fell like a bomb because it was developed by a small group of Chinese engineers with a handful of old GPUs, and it turned out to be superior than the best model that exists until then. Obviously, we need great machines to do great things, but with what we have, I think we can still get many things. That does not mean that we will need money to build Marenostrum 6, 7 and 8. We are going to see how far it comes to us, but I believe that Europe is willing to achieve the sovereignty of software and hardware. I think we will get it.
P. Another of the things that has happened with Trump is that many European scientists based there that consider return. Are you tantling someone?
R. I came back before this happened, therefore, I do not include myself directly, although since Trump won I saw things in the finish line that did not just convince me. For example, efforts were canceled in diversity and inclusion, issues I deal with. My colleagues from Google, Amazon or Tesla told you the same. There are people who feel uncomfortable in their work and students who will not be able to do their doctorates. If there is any particular profile that any of our teams consider important, we can speak it. We always look for the best talent.
P. A supercomputing center consumes a lot of energy. How does the BSC-CNS focus the environmental issue?
R. First, the 100% energy spent the Marenostrum 5 is 100% renewable. On the other hand, the building is really efficient, everything is designed to optimize energy. Obviously, we spend a lot of energy, with peaks of about 10 megawatts (MW), but it is also true that we have oscillations, you are not always using all the GPUs. I would like to emphasize that this is not a data center, we are not serving YouTube video traffic or processing Amazon purchases. Here we do science. I want to believe that the energy we consume is properly inverted.
