Elisenda Bou: "Artificial intelligence gets more people to have access to technology"

Elisenda Bou-Balust is co-founder and CTO of Vilynx, an emerging company with four industrial doctorates, at the forefront of creating the first artificial intelligence system that offers tagging and indexing services for content and videos on the internet, to be able to find them from searches. The company was Apple's first acquisition in Spain in 2020.
Image of the interview with Elisenda Bou

She is a telecommunications engineer from the Polytechnic University of Catalonia (UPC) and an electronic engineer from the University of Las Palmas de Gran Canaria (ULPGC). He did his doctorate at the UPC-MIT, and has collaborated with NASA , with the Massachusetts Institute of Technology ( MIT ) in Boston and with Google in the field of artificial intelligence applied to satellites. Very committed to education and STEM, she was co-founder of the UPC nanosatellite laboratory, where she has been teaching since 2016. She also often participates in initiatives such as Women in STEM in secondary and university education centers.

He has directed four industrial doctorates with Vilynx, with offices in Barcelona, New York and California, and has managed to establish this program for the first time at Apple. She has received many accolades such as the Young Entrepreneur Award from the Association of Young Entrepreneurs of Catalonia (Aijec), she has been awarded by Thales , GMV and Google and recently she was honored with the Fundació Princesa de Girona Empresa 2022 award.

 

First of all, congratulations on the Fundació Princesa de Girona Empresa 2022 award. The media talk about you as a global reference in the field of science and technology. What motivated you to become an engineer?

I was one of those girls who was very clear from the beginning that mathematics was my field: I was very interested in mathematics, engineering, technology and mechanics. What I wasn't sure about was which engineering, but I knew that engineering, for sure. Then I ended up in telecommunications because there was a bit of everything: a bit of electronics, programming, signal processing and antenna technology. All these things.

When you studied telecommunications there weren't many girls, I imagine.

There were very few. In my time there must have been three or four girls from my class, out of a class of one hundred. There are two types of girls. On the one hand, there are those girls like me, who are very clear that the stigma of being a geek is the same for them; but there are many girls, the majority of girls, who might do engineering; even so, they don't have such a clear vocation to think "it doesn't matter to me, technology is what I want to do, I'll do it no matter what." They feel like there's a bit of a stigma that because you're a girl and doing engineering or doing technology, you have to be a geek or a hacker or a weirdo.

It's a very silly prejudice. However, it is very present and, on the other hand, when the boys do not know what to do, they all go to engineering: because it has exits, because it will be good, because it is, etc. And this is somewhat the basic problem we have: yes, there are girls who, like me, are very clear about it and end up with it, but there are many who we lose along the way who could contribute a lot.

"There is the feeling that there is a certain stigma that, because you are a girl and do engineering or technology you must be a geek, hacker or weirdo, but that is not the case"

You just have to see how far you have come yourself. How did you experience the Fundació Princesa de Girona Empresa 2022 award? It is a great recognition of your career.

It is a very important recognition and I receive it with great humility and great joy. Because when you look at the awards from previous years or the women awarded this year, you feel very proud to be awarded. I think it's a really good recognition of all the work that the entire Vilynx team has done, which has been a lot for eight or nine years. We started very early, there was a lot of belief in this idea and people worked hard to make it happen. For me it is recognition for everyone.

What is behind the name Vilynx?

The name Vilynx is a bit strange, and comes from a well-known expression: "to have the sight of a lynx". It is a name composed of two words: "vi" which would come from sight and "lynx" for the animal. The idea is that if you have a lynx view, you'll be able to understand the content much better or see what's in the content better.

Your PhD was about artificial intelligence. Do you think the citizen is aware of its usefulness?

I think one of the good things about artificial intelligence is that it can go unnoticed. They somehow humanize machine interfaces. A very clear example is the car navigation, which many people might not use, but because it is a very simple voice system, everyone ends up using it. You address your navigation system as if it were an artificial person; artificial intelligence allows us such things. So, I think it's true that many people are not aware of the uses of artificial intelligence these days.

In a way I think it has this positive character because it naturalises interfaces, but without it being noticeable. And I think that's a really good thing, it's a beautiful thing that artificial intelligence has. You also need to keep in mind the whole issue of accessibility. There are so many uses of artificial intelligence in this field that are also very important. In short, it allows us to make technology accessible to more people, and more diverse people.

"In a way, artificial intelligence humanizes machine interfaces"

Image of the interview with Elisenda Bou

Has the AI research you did during your PhD become obsolete?

The speed with which artificial intelligence systems become obsolete is very great, because we have so many people working at the same time and that is a very good thing. Perhaps, saying that it is obsolete would be like saying that it is no longer useful. It's always moving forward, it's a fact.

While doing a PhD or industrial PhD, you evolve in a line of knowledge that humanity has, but also important is the evolution you make as a person and the resources or methodologies you learn, which you will then continue to use for the rest of your life

Let's talk about our house, now. Do you consider Barcelona a point of reference in the field of artificial intelligence?

Yes, I believe that Barcelona is a benchmark in the field of artificial intelligence , and it is so for many reasons. Historically, we've had a very good education here in communications engineering, which is the basis of many artificial intelligence systems. The UPC and the UPF have highly renowned experts in artificial intelligence, and we are seeing how a knowledge center is being created around the city. This can be seen both in the migration of talent returning to Barcelona and in those coming to Barcelona for the first time to work in this field. We also see this in the number of startups we have that use different AI systems.

We have a very large pool of talent thanks to our entire university network, which is very powerful. At the same time, we are at a very nice point in the ecosystem, where there are many start-ups and new projects in artificial intelligence that are very successful.

"Barcelona has a very large pool of talent in artificial intelligence thanks to the powerful university network we have"

Awards such as the Princesa de Girona distinguish emerging young talent. How important is this talent in the industrial PhD projects you have led?
I think it's the most crucial. It is obvious that without talent you can do absolutely nothing. At Vilynx, one of the things we did was to have the first offices within the UPC campus, to attract talent as easily as possible and to have direct access to them. There is no innovation without talent, and not only on the innovation side, but also on the execution side. Vilynx was a project made by and for a team.

Do you think industrial doctorates are a good tool for training this talent?
It's great, because as a company you always aspire to attract the best talent possible, and retain them in some way: that they are so good that they don't want to leave. But when you have very powerful talent in the academic field, it is very normal for these people to want to pursue their vocation at higher levels. They take a job at a start-up company and after a few years think that maybe their career advancement involves getting a Ph.D. Then the opportunity that industrial doctorates provide is to allow them to work in the company and at the same time allow themselves to have this recognition at the doctorate level, and continue to advance in their academic career. I think it's a luxury to have this program, and it's wonderful for retaining talent, but also for workers who want to continue their careers undecided whether to stay in industry or go into academia.

Jorge Wagensberg said that "innovation requires three things: having a good idea, realizing that it is good and convincing others, and it is almost never the same person who achieves all three". After leading four industrial PhD projects, do you agree with this statement?

Totally agree. You don't just have to convince, then you have to execute the idea. In all cases, you will definitely need a set of people with different abilities and strengths. Absolutely agree with this sentence of Wagensberg . I think that everything that is done in innovation processes is never the work of one; when we really see excellence and see results is when we have teams focused on this innovation. We are talking about an industrial doctorate in a context: a research team, a team at the university and in the company, a support, etc. It is here that we see the breeding ground is created for ideas to flow and then also be executed. It's never just one person.

It's a paradigm shift. We have passed the age of inventors, now to innovate you need to collaborate.

This happens for two reasons. The first is that the complexity of the systems has increased enormously. Now it is impossible to find someone who understands a whole complex system. The second reason, which is also important, is specialization. Each subsystem is part of a larger and increasingly specialized system. That's why, if you want to innovate, you need very specialized talent.

What do the four projects you have led in common?

They are different projects. We have one on natural language thinking, on computer vision and we also have automatic voice recognition projects. They are different from each other, although they are also part of a complex system, as we said before.

Projects to automatically analyze, index and tag video content. What does this technology consist of?

The core technology we made at Vilynx was self-learning systems. When we started, most research groups and companies were doing supervised learning. What does this mean? Basically, you create a dataset and tell it, for example, that you want to recognize cats and dogs; then the artificial intelligence system takes images from the data set and then says it's a cat or it's a dog. But when an image of a zebra arrives, which is neither a cat nor a dog, the system doesn't know what to do. What we were doing at Vilynx was unsupervised autonomous learning: the zebra is neither, but it looks like it also has four legs, it also has hair, and it has some characteristics that are similar. The artificial intelligence won't give it a name because it doesn't know it; even so, he is able to distinguish the zebra from a cat and a dog he already knows.

Can all internet videos be indexed?

They can be indexed, that's all! This is precisely what we intended at Vilynx.

And in the future, how do you think information will be organized?
We always start from a duality: humans work with unstructured data and computers work with structured data. There are more and more artificial intelligence systems that help structure data to solve doubts, to be able to search for things at home, etc. But these systems are also dedicated to re-normalizing structured data so that the interface is as human as possible.

"There are more and more artificial intelligence systems that help structure data to resolve doubts"

Will we always have this duality between humans and machines?

That's because we humans communicate in an unstructured way, and we don't yet have any software that literally works like that. That is, first we save it in the system as values and variables and then we translate it back into human language.

How do machines learn?

When we talk about learning, I think it's important to make visible that what we're trying to do is get machines to have naturalized behaviors. What does deep learning enable us to do? Well, to do precisely the translation we said before between structured and unstructured language.

How has the experience been with industrial doctorates?

I am a strong supporter of these projects, and I am delighted with the Industrial Doctorate Program, because for Vilynx it was a turning point. We were at a point where we were fortunate to have a very leading team. But we also saw that, for the team to continue to be happy doing what they were doing, we needed to offer an exit to be able to do a PhD. It has been very enriching for us. It has helped us a lot in retaining talent, and also in making sure that talent is comfortable and able to pursue their academic training. They have also created a very beautiful dynamic of innovation and research. Although the base of companies like Vilynx is R+D+I, industrial PhD projects have strengthened this base, helping to establish a commitment to this kind of research. It has been very enriching for everyone.