ProLOEWE faces

Caption: Constantin Rothkopf (left) and Kristian Kersting (right), both professors at TU Darmstadt, have been working together since their first meeting in Osnabrück, where they discussed research ideas in the bar-café “Tiefenrausch“ (translated roughly as "raptures of the deep") after a long day at the university. It was there that the first ideas for a joint (LOEWE) project emerged.

Professor Kersting, Professor Rothkopf, the current issue of ProLOEWE faces will be the first double portrait. The reason for this will become clear to our readers as soon as they learn more about the origins of LOEWE-WhiteBox, for which you both serve as spokespersons. What is the goal you want to achieve with LOEWE-WhiteBox, and what makes your collaboration on this topic so special and inseparable?

[Constantin A. Rothkopf] With LOEWE-WhiteBox, we want to ensure that artificial intelligence no longer remains an impenetrable “black box”, but is designed in such a way that we can understand and trust its decisions. In many areas, it’s not enough for a system to simply deliver correct results; we also need to be able to understand how these results are generated. Imagine, for example, a medical diagnostic system that recommends a treatment. For doctors and patients, it’s not enough to know that “the computer says: this is how it is.” They want to understand the factors that led to this recommendation. This is precisely where WhiteBox comes in.

What’s special about our project is the close collaboration between two disciplines that are rarely combined so consistently: artificial intelligence (AI) and cognitive science. AI brings high-performance methods that can process millions of data in a short time, but are difficult to understand. Cognitive science, on the other hand, has been asking for decades: How do we actually explain intelligent behavior in humans? We humans, too, are a kind of black box – no one understands the brain in all its details. But we have found ways to make behavior understandable at a higher level, for example, by quantitatively describing goals, strategies, or rules and explaining them computationally.

For example, when a child learns to avoid obstacles while riding a bicycle, we can’t explain it by the wiring of its billions of neurons. But we can say: “The child looks ahead, recognizes the obstacle, and decides to steer left.” This kind of explanation is plausible. Of course, in cognitive science, we don’t only have verbal descriptions; we can also formulate such explanations mathematically and algorithmically. The goal of WhiteBox is to transfer cognitive explanations to AI so that machines can also be explained in the way we humans expect it of each other.

It is precisely this connection, the power of modern AI and the explanatory power of cognitive science, that makes our project unique. Only when both disciplines work together can we develop AI that is not only powerful, but also transparent, trustworthy, and “human-centered” in the best sense of the word.

Many people may have a somewhat idealized vision of what it means to be a professor. Can you tell us something about this based on your own career path?

[Kristian Kersting] Many people imagine the daily life of a professor as one where they can calmly reflect on big ideas, perhaps discuss them with students, and thus advance science. In reality, the path to this, and even the daily routine itself, is much more complex – and often challenging.

My own career path demonstrates this: There is no direct, straightforward path to a professorship. You go through various stages at home and abroad, have to assert yourself against international competition and invest a great deal of time in proposals and reviews. This contradiction is particularly evident in work on artificial intelligence: Research is developing exponentially fast, but our funding system is linear and slow. While we should actually be implementing new ideas quickly, we lose a lot of time in lengthy processes. The image that professors and professors primarily spend time thinking is therefore unfortunately not true.

Added to this are challenges within the science system itself:research funding is scarce, and when new resources flow into a groundbreaking field like AI, it often means that other areas receive less. This leads to battles over distribution and sometimes even envy – things that are rarely seen in public, but are unfortunately part of the story. At the same time, the world isn’t waiting for German professors – it’s waiting for solutions. If we hesitate, other countries may soon set the course.

This makes successes like LOEWE-WhiteBox and the two Clusters of Excellence “The Adaptive Mind” and “Reasonable AI”, which were created in part thanks to WhiteBox, all the more important. They are not only scientific beacons, but also proof that we can create internationally visible structures that promote research and young talent alike. For me personally, these projects are a true success story: We only began collaborating in these fields at TU Darmstadt nine years ago. They give us hope that, especially in these two Clusters of Excellence, we as scientists can once again take time to think – time that we urgently need to help shape the next major step in AI development from Hesse.

After describing the challenges of academic life, what is special about the LOEWE program?

[Kristian Kersting] The LOEWE program is something special in Germany because it can create freedom. Many large national or European programs are more drawn out and lack regional relevance, so innovative ideas can get lost in the mass of applications. LOEWE, on the other hand, deliberately focuses on Hessian funding: fast, direct, close to researchers, and with a clear focus on its own state. Almost like a savings bank. When applying for a loan from a large international bank, the process is often very lengthy, bureaucratic, and impersonal. A local savings bank is much more targeted because it knows the region and understands exactly where the strengths and needs lie. For us at WhiteBox, this freedom was crucial. We were able to build a bridge between artificial intelligence and cognitive science, something we believed in from the start, but which might have been considered too risky or too unusual in traditional programs. LOEWE enabled us to be bold and work across disciplines.

This is also a major advantage for Hesse: LOEWE specifically promotes the region’s strengths and raises their international profile. WhiteBox, as well as the resulting Clusters of Excellence “The Adaptive Mind” and “Reasonable AI,” clearly demonstrate how regional funding can create true beacons that radiate far beyond Hesse.

In short: LOEWE gives us the freedom to think scientifically and explore new approaches – and at the same time, it contributes to Hesse’s profile as a location for cutting-edge research sustainably.

Now the funding for your LOEWE project is running out at the end of this year, but with RAI and TAM, you have managed to secure and participate in two Clusters of Excellence. A truly extraordinary achievement! Please explain to us why this is so important, what it means to work toward excellence, and, of course, how much LOEWE is involved.

[Constantin A. Rothkopf] The fact that we were able to establish two Clusters of Excellence from WhiteBox, “The Adaptive Mind” and “Reasonable AI”, is indeed an extraordinary success, not only for us researchers, but for the state of Hesse as a whole. Excellence doesn’t mean promising a revolution every year. Excellence means setting the pace internationally and not just reacting. It means the highest scientific quality, critically questioning trends, and the freedom to develop new, unconventional ideas. These are precisely the conditions under which real breakthroughs can emerge.

This is of great importance for a state like Hesse: Clusters of Excellence are not just scientific flagships, they also attract talent, companies, and investments. They create an environment where the best minds stay or come to us because they know they can conduct world-class research here. This creates an innovation eco- system that extends far beyond the universities into the economy and society. Support for startups cannot be considered in isolation; the ecosystem must be taken into account as well.

And, of course, LOEWE has a lot to do with this success. Without targeted, regional funding, we probably wouldn’t have been able to realize the risky idea of systematically merging AI and cognitive science. LOEWE gave us the freedom to dare to try new things. This freedom has given rise to structures that are now internationally visible. We can say: LOEWE was and remains a key building block for the Clusters of Excellence and is a part of the foundation upon which Hesse today plays an internationally visible role in artificial intelligence and cognitive science. This makes it all the more important that Hesse continues to support science!

How do you manage to switch off from (scientific) work and have time for your private life?

[Kristian Kersting] To be honest, truly switching off from work is often difficult in science, especially in (artificial) intelligence research, which is developing so rapidly. Many people think you can simply delegate proposals and concepts. But that's not true. We write them ourselves because we bear the responsibility and lead the way. That takes time – time that you often can't spend with family and friends.

This makes the trust and support we receive from those around us all the more important to us. Family and close friends know that we're not always available, yet they support us nonetheless. This isn't something we can take for granted, and we're very grateful for it. Without this support, this work wouldn't be possible.

* Translated roughly as "raptures of the deep".

The interview was conducted by Tanja Desch.