You are an expert in machine learning, especially in kernel methods or causality. What work has he done in this area?

Kernel methods are a field of machine learning. In fact, they are methods of learning standards, addictions or relationships from observations, and they are special because they are connected to different areas of mathematics: functional analysis, optimization theory… In this way, they provide a way to learn norms from data, nice and elegant, which can be used in fields like medicine or industry. They are useful not only to know the relationships between the data, but to go further and detect causality. So it goes a deeper level than machine learning.

One of the applications is perhaps not the first to think. It has been used to detect exo­ple­net. How was that?

Well. I was spending a sabbatical year with my family in New York and I started talking to astronomers. They told me that from the data collected by the Kepler space telescope, they had a problem detecting exoplanets. Because to detect exoplanets, as they passed past their star, they were based on the loss of light from it. But this loss is very small and it is difficult to distinguish it from the noise produced by observation itself. This noise and telescope errors can be deduced from data from other stars and we leverage it.

In a few years' time, we found between 20 and 30 copies, which were subsequently confirmed by other methods. And soon, it was confirmed that one of them was at a vivible distance from his star. This means it was at a distance that allowed for liquid water. The first exoplanet detected was exhilarating.

In what other areas is it applicable?

In principle, in all areas where experimental observations can be made. It is necessary, on the one hand, to have a large amount of data, many measurements and, on the other, to establish relationships between different observations. For example, we can relate the measurements of a biomarker to the risk of a tumor and know if this biomarker can predict a tumor before detecting it. That is, a standard can be deduced from the data, even when it is difficult for a person to perceive it.

Apart from Kernel methods, what do you think will be the main developments of artificial intelligence?

Right now, one of the most exciting areas is natural language processing (NLP). 20 or 10 years ago, numerous investigations were conducted on computer vision: how to recognize objects, or how to separate cars and pedestrians… to develop automatic cars, for example. Much progress has been made in language analysis and text forecasting over the past 10 years. And suddenly programs are able to produce texts. This has generated astonishment and admiration, and I think it will have a great development in the coming years.

 

Ed. © MPI for Intelligent Systems / Annette Cardinale

How can we ensure that the benefits of AI systems reach everyone?

To make the benefits of AI systems as accessible as possible, the first step is to publish research. Not only describe methods in a scientific journal, but also publish the code. What's more, the data that has been used to train this code.

However, this is not just any question. For example, in medicine, data can't be made public in any way. Or, in another area, it may not be appropriate, as it risks being malicious. But in general, I believe that if we want it to be beneficial to the largest number of people and societies, research must be as transparent as possible.

Can artificial intelligence create problems in other areas like the environment?

It is true that training artificial intelligence programs requires enormous energy that consumes a lot of resources. And research centers are not always in the most appropriate place from the perspective of sustainability either.

But from another point of view, artificial intelligence can also have a positive effect on the environment. Indeed, climate is a complex system for which we do not have a complete model. Using artificial intelligences can also help us understand climate, which will help us slow or mitigate climate change. It will also help us create better technologies: better solar cells, better electric batteries…

I have read that he has refused to participate in military investigations. Is that the case?

Yes. Traditionally, man is involved in wars. Someone decides to attack the other or shoot him, someone orders and it's someone's decision. This means that there are responsibilities and moral judgments.

And war is always bad. But if machines get smarter and more responsible for making those decisions, who is the responsibility? Who has developed a program to know and kill a person? What have you designed? From the seller? The government using the system? It is difficult to foresee how the war can change.

I believe that we must discuss this and that this is an issue that needs to be looked at very carefully. But I'm very skeptical, I don't see that we're at it. I therefore always try to protect and support those who say that such weapons must be banned.

On the other hand, you've learned philosophy. Do you think it is useful or, moreover, it is necessary to consider philosophy in the formation of artificial intelligence?

Certainly, philosophy has been very useful to me. I think anything you learn can be useful once. And if you've learned philosophy, it will influence how you think about the problems of artificial intelligence. Of course, many technicians only deal with the technical aspects of artificial intelligence. But the questions that we want to clarify are not just technical, that is, how are we going to detect the structure of the world? Why does the world seem to be subject to law and not to chance? In my opinion, these kinds of questions, in depth, are philosophical and from that perspective we must understand them.

What other disciplines should be taken into account?

In short, it will influence when it comes to artificial intelligence, so all aspects, from natural sciences to art, should be taken into account. But economics, politics, society -- indeed, society needs more debate.

 

Ed. FBBVA

What do you think of requests for temporary cessation of the investigation?

As I have just said, I think it is necessary to discuss and reflect on the development and consequences of artificial intelligence. But these moratoriums are not realistic, it won't happen. It is significant that behind one of the requests is Elon Ms. The truth is that there is competition between companies, and those who come from behind are asking for time to catch their predecessors.

It is much more important to understand how these systems work, what consequences they can have and how we can ensure that they are used for the benefit. And you have to do that with people, you can't leave everything in the hands of companies.

What can artificial intelligence bring in the future now?

Question not answered. We cannot foresee what will happen. Even when the Internet was created, it was impossible to know how we would use it today and what it was going to bring us. We currently have systems capable of creating poems. There is a leap forward and in the future the conversation between machines and people will become more and more common.

We need to think about how to strengthen society to be aware of the good and bad use of artificial intelligence.