At our two locations in Dresden and Leipzig, we investigate the need of AI applications for high quality data and formalized knowledge to achieve valid and reliable prediction and analytical results. We therefore combine research on new Fundamental AI methods not only with our Big Data research on data integration and data quality, but also with new methods for data acquisition and visualization to support data-driven AI. In addition, AI methods need to be systematically integrated into scientific analysis workflows, which can accelerate research progress in many areas. In addition, trust, transparency, and traceability of AI-driven decisions and processes are key. Finally, privacy and informational self-determination remain largely unresolved issues that we will tackle with research on privacy-preserving machine learning.

We contribute to the future development and use of AI in five key areas:

Applied AI and Big Data

In addition to methodological and fundamental research, it is our ambition to apply AI and Big Data methods in an interdisciplinary way, thus advancing research excellence and progress in general. For this reason, scientists from a wide range of application fields are working on making methods of AI-based data processing, knowledge representation and evaluation usable for the fields of medicine, natural sciences or humanities and social sciences.

AI Algorithms and Methods

We work with new mathematical methods to drive AI applications in various domains. The development of new approaches to graph-based AI as well as knowledge representation are at the focus of our research. We aim to make algorithmic AI decisions and processes explainable to promote trust and acceptance. Furthermore, distributed learning of very large models, further development of natural language processing models, and the investigation and development of neuromorph-inspired architectures are important research topics.

Big Data Analytics and Engineering

Large data sets from social networks, multimedia collections or scientific experiments and their analysis enable a variety of new options. We are developing methods and solutions to investigate large and complex data sets from various application areas in science and industry. Our research is based on the observation that Big Data solutions can only be developed by adapting the entire data lifecycle as well as by accessing modern data processing architectures.

Responsible AI: Ethical and Societal Dimensions

Big Data and AI technologies are becoming increasingly significant in the lives of both individuals and society. However, the omnipresence of these technologies harbors numerous uncertainties or even risks in terms of data protection and data security. That is why, in addition to the further development of machine learning, our scientists are also working on new techniques for preserving privacy in the application of these methods. Furthermore, the explainability of AI decisions is being investigated in order to make them transparent and understandable thus strengthening trust in AI methods in the long term. Legal and ethical issues are also being researched in this context.

Architectures / Scalability / Security

In order to create a foundation for the use and research of new AI methods and applications, the long-term goal of ScaDS.AI Dresden/Leipzig is to provide flexible, scalable and secure computing solutions. The basis for this is created with the preparation of high-quality training data. The focus is on scalable, agile and secure software architectures as well as online workflows. In this context, the term architecture describes the entire computing platform, both from a hardware and software perspective, from large-scale computing systems to nodal and component architectures. The technologies emerging at the center, so-called hardware/software computing platforms, are then made available to a variety of applications and users.