Urs Bergmann has to think for a moment. After all, it’s been several years since he began to take an interest in artificial intelligence. "I was 13," he begins, "when I began to read the first books on neural networks.” His interest soon became more: a vocation.
This led him to Zalando, where he heads two teams in the research department. One is the Advanced Image Manipulation team. "Put simply, it's about machines learning to generate new images similar to the first ones based on input data; a first step towards machine creativity," he says. His other area of expertise is Intelligent Control. Digital systems learn to deal with the environment independently, i.e. they develop and test hypotheses. For example, these systems can get to know customers better than classical systems in order to suggest better recommendations for articles that are more interesting to them.
Urs, where does your connection to artificial intelligence come from?
On a smaller and larger scale, I have been involved with this topic for a very long time. I studied physics and had already dabbled in computer science before. Eventually, I wrote my doctoral thesis on neural networks in brain research. I have actually always approached the topic from the biological side. With Zalando, I now mainly take the engineer's perspective and research mainly in fields of application.
How did you develop an affinity for the topic?
My older brother introduced me to neural networks many years ago. I was still a teenager and didn't have the necessary mathematical tools for a deeper understanding, but I found it totally thrilling. I've always liked creating worlds. As a teenager, for example, I programmed computer games. Artificial intelligence is then, so to speak, one step further: you create something that can find its way around worlds.
How important is the topic for Zalando?
Artificial intelligence is one of Zalando's current focal topics, which is pursued on several fronts, from infrastructure to research in an application-oriented, academic framework. For example, about 120 researchers work at Zalando in the field of machine learning.
In which fields are we working?
Altogether, the fields can be divided into two major subject areas. First: Business optimization. This includes optimization processes in logistics or fraud detection. Secondly, a large number of products are aimed at personal relevance for the customer. These include article recommendation systems or size recommendations, but also personalization of the shopping experience, search, or an outfit recommendation system.
Artificial intelligence is one of Zalando's current focal topics, which is pursued on several fronts, from infrastructure to research in an application-oriented, academic framework.
You are currently a member of an expert group of the European Commission discussing and thinking ahead about Artificial Intelligence in the EU. What exactly is the aim of the expert group?
Representatives from academia (science, social sciences, and the humanities), civil society, and industry come together. The aim is to advance the topic of artificial intelligence in Europe and to give direct recommendations for the future - for example in terms of laws and policy, but also to develop ethical guidelines for this.
Why is Zalando a member of this group of experts?
I think because, as a European company, we are almost an exception. We have grown up very quickly on the Internet and are also relying on artificial intelligence. There are not many others in Europe in comparison, and we can contribute our expertise to further development.
The European Commission is committed to building trust among citizens through ethical guidelines and a European understanding of AI.
What exactly happens at these expert meetings?
In principle, there are two main areas of work. One deals with ethical questions, the other with questions of policy and investments on the part of the EU. The European Commission is committed to building trust among citizens through ethical guidelines and a European understanding of AI. The application of AI on the basis of common European principles and values can become a differentiating feature for the EU. A first draft was published on the 18th of December and was open for feedback until beginning of February. The investments, in turn, are to be massively expanded. The question that will also be discussed is how the greatest possible output can be achieved. Both areas of work contribute to making the EU more competitive.
Because the EU has a lot of catching up to do compared with China and the USA?
At the moment, companies from the USA are clearly the leaders. But China, too, is investing a tremendous amount. They have a plan to be world market leaders in some technological areas by 2030, including AI. In Europe, we have many excellent scientists in artificial intelligence. One problem is that many of them are poached by non-European companies.
Some say that Europe has already been left behind. What do you say to that?
Perhaps it is true in the area of deep learning, with the exception of a few research groups. But there are many areas of AI in which we are well-positioned. In particular, the combination of AI with, for example, our strong machine and vehicle industry promises great potential. A major challenge, however, is the use of the methods; after all, medium-sized companies cannot afford large teams of data scientists in the same way technology companies do. Even DAX companies sometimes find it difficult to attract sufficient highly qualified personnel in this area. It’s concerning that the German AI Strategy, which was made public last year, won’t change much. It has too little focus on neural networks and deep learning, and these are the areas in which we are seeing the most exciting progress.
What path should Europe take to further advance artificial intelligence?
Since the AI currently used is mainly based on deep learning, this area especially should be pushed, I would say. Methods from this area mainly require three factors. First, a lot of data. Second, a lot of computing capacity on special processors. And thirdly, well-trained personnel. Building on this, Europe should find ways to promote all three factors, e.g. through training programmes, immigration laws, or making large and high-quality data accessible.
And the expert group of the European Commission can initiate this?
I believe that it will certainly have an impact. The group has been set up by the European Commission and covers a wide range of society's opinions. The European Commission has also recognized the need for action. There is one thing on which everyone basically agrees: In future, almost no European industry can remain competitive without the use of artificial intelligence.