At present it is frequently the case that the boom in artificial intelligence (AI) is being curbed in Germany by a shortage of highly qualified young experts. As one of three new Zuse Schools, the Konrad Zuse School of Excellence in Reliable AI (relAI), a joint project run by the Technical University of Munich (TUM) and Ludwig-Maximilians-Universität München (LMU), is intended to counter this trend. Co-spokesperson Gitta Kutyniok, a professor of the mathematical foundations of AI at LMU, thus sees the support provided by the German Academic Exchange Service (DAAD) as a milestone. “It is helping us make significant progress in the fierce competition for the world’s best AI students.” The school, which the DAAD is providing with a total of 13.8 million euros in funding courtesy of Germany’s Federal Ministry of Education and Research, is focusing its activities on the key issue of AI’s reliability. How can it be guaranteed that self-learning algorithms arrive at the right decisions because they have learnt and deduced correctly? Working with around 20 partners in research and industry, relAI explores fundamental questions relating to the application of AI technologies.

 

INTERVIEW

Self-learning algorithms: how reliable are decisions taken by AI?

TUM is involved in two of the three Zuse Schools. Professor Stephan Günnemann, the school’s spokesperson and a professor of data analytics and machine learning at TUM, explains how important cooperation on the subject of AI reliability is, both at the local level and with international partners.

Professor Günnemann, what sets relAI apart as an excellent graduate school for budding AI experts?

In focusing on the reliability of AI technology, the project partners TUM and LMU are addressing a highly relevant topic that is not only of technological interest. It also has a strong social impact, as people will not wish to use AI methods in society unless they are secure and reliable. This focus is decisive, and what sets us apart.

How do you research the reliability of AI?

On the one hand, we combine basic and methodological research with direct application in relevant domains, such as robotics, driverless vehicles, medicine and decision-making. We work with a broad spectrum of researchers – mathematicians and computer scientists, but also experts from the different areas of application. We also attach importance to this interdisciplinary collaboration during training, our doctoral students being supervised by experts from both method-based and application-oriented research. In the medium term, we are also planning to cooperate closely with industrial partners so that the methods developed at our school can be trialled in companies.

You yourself have worked in the research department of a company, so you also know the other side. Does a lot need to be done in Germany to bring science and business closer together?

AI is a good area in which to work closely together, as there is always a concrete objective. Industry has a good knowledge of the real-life problems, which makes cooperation particularly productive. At relAI, it will also be a question of bringing master’s students into contact with firms at an early stage. After all, the goal of the Zuse Schools is not only to convey practical experience in close exchange with business. Candidates are also to be given the chance to discover everything that can be done in the area of AI – many students are not aware of all the potential applications.

TUM has been receiving the DAAD funding since July 2022. What are the next steps?

We have already staged our first Fellow Assembly, which brought all the founding members together and also saw the boards formally elected. Currently we are busy fine-tuning the curriculum and consolidating the teaching that will be offered during the upcoming semester. One major advantage is that the graduate school is embedded in TUM’s Munich Data Science Institute, which is helping us set up the Zuse School until we have our own team in place. In parallel with the organisational activities, we now need to raise the Zuse School’s international profile and ensure that we can attract the right candidates.

What do you hope to gain in this context from your cooperation with the DAAD?

The DAAD is known worldwide for facilitating exchange between students and researchers, and has an established network of strong partners.  We believe that this will ensure a high profile for our Zuse School and allow us to recruit international talents.

The two partner universities, TUM and LMU, work together with both regional and highly prominent international partners in the Zuse School.

The network of cooperation partners is wide-ranging. It encompasses not only the Fraunhofer Institute for Cognitive Systems, the Fraunhofer Institute for Applied and Integrated Security, and Helmholtz Munich, but also AI centres from around the world, for instance at the universities of Stanford and Princeton. AI is a topic of global relevance and the research community is international. Exchange with the world’s leading universities is thus a key element within the school. Pursuing excellent research means networking internationally. Our participants are to be given the opportunity to do research stays with our partners. At the same time, we support international fellows who come to Germany and work as visiting scholars at the school, either doing research work or giving lecture series.

TUM is involved in two of the Zuse Schools – are there points of overlap, do you engage in exchange?

Even during the application phase it was always our goal to work together with the two other schools, jointly running workshops or fairs. TUM is also involved in the ELIZA School at TU Darmstadt, which likewise focuses on machine learning. We cooperate closely with our Munich colleagues from this department in any case, so there are bound to be aspects we can follow up on. Ultimately, however, it is not only about joint activities, but above all about making the schools generally known. Our joint objective is to perform on the international stage with these schools, promoting the advantages of Germany as a land of innovation. This is what matters to us, which is also why collaboration is important.

And what is the specific objective of the cooperation partners with respect to relAI?

The visibility of this topic and the success of participants are the top priority for us. On the one hand, we hope that Germany will acquire an international reputation for reliable AI, and that students will come to the school in Munich to pursue research in this field. For me, the second goal is for the participants to genuinely benefit from the school, to be successful, to establish a network thanks to all the partners we have on board, and to remain in contact with one another. To this end, we are also planning to set up an alumni network in the longer run. Successful graduates who work at top companies or become professors at good universities and are later willing themselves to contribute to the school have important boost potential. All of these are effects we want to achieve with the school.

 you can find more information on relAI.

Interview: Gunda Achterhold

Networks and collaborations