back to list

Strengthening basic AI research

DFKI establishes laboratory at TU Darmstadt

2022/06/09

The German Research Center for Artificial Intelligence (DFKI) is expanding basic AI research with a laboratory at TU Darmstadt. The focus is on Systemic AI, the combination of different AI building blocks into complex, comprehensive AI systems.

Professor Carsten Binnig, Professor Jan Peters and Professor Kristian Kersting (from left).

“The integration into the DFKI network strengthens research and application of artificial intelligence throughout Hessen,” explains Hesse's Science Minister Angela Dorn. “In particular, the close cooperation with the state-funded Hessian Center for Artificial Intelligence (hessian.AI), which involves 13 Hessian universities of all types, will act as a catalyst. I believe that everyone involved will benefit sustainably from this new facility, the scientists and students of TU Darmstadt and the other universities involved in hessian.AI, but also business and society. This is because DFKI cooperates with a large number of national and international companies, thus strengthening the transfer of scientific knowledge.”

The state of Hesse plans to fund the establishment of the new DFKI laboratory with six million euros over the next three years. A corresponding letter of intent was signed at the end of 2021; the first tranche of two million euros has now been approved.

The establishment of a laboratory of the German Research Center for Artificial Intelligence at TU Darmstadt is a milestone in the establishment of the Darmstadt AI ecosystem, which I am very pleased about. With DFKI, we gain another excellent non-university research institution as an important partner to further expand the comprehensive network at TU Darmstadt. I am convinced that with the proven expertise of the colleagues involved, we will be able to further intensify basic research at DFKI.

Professorin Dr. Tanja Brühl, President of TU Darmstadt

Professor Dr. Antonio Krüger, Chairman of the DFKI Executive Board: “With the DFKI Laboratory Darmstadt, we are specifically strengthening basic research in the field of machine learning. We are gaining internationally highly renowned colleagues with an excellently networked environment, whose systemic approach fits perfectly with DFKI's holistic AI philosophy. We are looking forward to the new impulses for our shareholders and partners, for academic activities, and in the medium term also for the transfer between basic research and application in industry and society. In synergy with the other DFKI Research Departments, we want to play a major role in shaping the third wave of AI, based on scientific excellence, from Germany and Europe.”

Three new research departments at TU

The DFKI is establishing a new laboratory in Darmstadt with three new research departments. The directors of the three departments of the DFKI Laboratory Darmstadt are professors of the Department of Computer Science at the TU Darmstadt, who have a strong track record in Artificial Intelligence and who are co-founders of hessian.AI, the Hessian Center for Artificial Intelligence.

“Systems AI for Robot Learning” (SAIROL) is the focus of Professor Jan Peters, who is also site director of the DFKI Laboratory Darmstadt. The SAIROL researchers will devote themselves to all questions relating to machine learning and robotics, drawing on the expertise of Professor Peters‘ group at the TU.

Professor Dr. Jan Peters, Head of the Intelligent Autonomous Systems Group at TU:
“The DFKI Laboratory Darmstadt will strengthen the basic research at the DFKI. My research area, the SAIROL Laboratory, focuses on basic research into machine learning for intelligent autonomous robot systems. This includes the development of methods for ”Systems AI“ for robot learning, through which robots develop new motor skills and problem-solving strategies from the interaction with their environment and human teachers.”

Under the direction of Professor Kristian Kersting, the “Systems AI” group (SAINT) is developing the algorithmic and programmatic foundations of the next generation of AI, which considers, understands and uses the interaction of individual AI algorithms as building blocks for more complex AI system in a mathematically and computationally sound manner.

Professor Dr. Kristian Kersting, Head of the AI and Machine Learning Lab at TU:
“At the heart of our research area is the question of how to combine low-level perception and high-level reasoning that are typically addressed using disparate AI approaches. While low-level perception (e.g., for recognizing a street sign) is handled by standard machine learning techniques, most prominently deep learning, high-level reasoning (e.g., to conclude that a street sign of 120 km/h does not make sense in a small village) is captured by using logical and probabilistic representations. To achieve this, we need to rethink AI from the ground up and lay the algorithmic foundations for ”Systems AI.“ Systems AI captures, understands and uses the interactions among AI building blocks (algorithms) to describe a single, complex AI system. The ecosystem consisting of TU Darmstadt, hessian.AI and DFKI is the ideal place to achieve this grand vision of AI.”

The research area “Systems AI for Decision Support” (SAIDE) is directed by Professor Carsten Binnig. The SAIDE Group is working on the next generation of enterprise information systems that support users with far more extensive AI-enabled functions than before. The automation of decision-making through systemic AI methods plays a central role here.

Professor Dr. Carsten Binnig, head of the Data Management Group at TU:
“Data-driven decision making today requires to manually design complex data transformation and analysis pipelines, which is not only a highly time-intensive task but also requires skilled IT-expert who are a scarce resource. With the research group “Systems AI for Decision Support” we are focusing on the foundations of how we can automate the design of such pipelines by developing new Systems-AI methods. Overall, this will not only enable faster decision making but also help to democratise data-driven decision making since non IT-experts can directly gain insights from data without involving IT-experts in the first place.”