Overview to AI-Research
The national competence center for big data, ScaDS Dresden/Leipzig, which exists since 2014, is being expanded to become one of the German centers for artificial intelligence (AI), which is funded as part of the federal government's AI strategy. This expanded center is called ScaDS.AI (Center for Scalable Data Analytics and Artificial Intelligence) Dresden/Leipzig. The project is funded by the Federal Ministry of Education and Research and is to be further strengthened by the Free State of Saxony with the establishment of several new AI professorships at both locations. In basic research on AI methods, the center strives to bridge the gap between the efficient use of mass data, advanced AI methods and knowledge management. In addition to new methods of machine learning and artificial intelligence, the focus is also on research topics on trust, protection of privacy, transparency, protection of minorities and traceability of AI-driven decisions.
The research is running at two locations, Dresden and Leipzig, by the partners Dresden University of Technology, Leipzig University, Max Planck Institute for Molecular Cell Biology and Genetics, Leibniz Institute for Ecological Spatial Planning, Helmholtz Center for Environmental Research, Leipzig and the Helmholtz Center Dresden Rossendorf.
In contrast to classical machine learning, artificial intelligence (AI) intends to solve problems, identify patterns, interact with users and is able to perceive and comprehend. In order to do so, AI requires access to high quality data and formalized knowledge. With the combination of research on knowledge acquisition, representation and foundational research on AI methods, we are certain to make significant progress in knowledge-driven artificial intelligence methods in next three years and beyond. Moreover, AI methods need to be systematically embedded into scientific analysis workflows which can speed-up research progress in many other research fields. Data analysis increasingly requires highly interactive and iterative data driven workflows of trial and error, inspecting intermediate results and adapting the analysis in a closed loop fashion. In business, AI needs to be embedded into the creation of product designs, services and business models. There is also a strong need for trust, transparency, and traceability of AI-driven decisions and processes. Finally, preserving rights of privacy and informational self-determination of citizens is still a largely unsolved issue.
Therefore, to foster AI developments and their further improvement ScaDS.AI plans research work in several important topics and wants to bridge the gap between efficient use of mass data, advanced AI methods and knowledge representation:
- ScaDS.AI investigates crucial foundations of provisioning high-quality data from multiple sources with the help of machine learning, which involves enriching data with background knowledge to build large reliable knowledge bases that are crucial in the development of reliable artificial intelligence models.
- ScaDS.AI intends to further focus its research on expressive data representations such as graph data to build AI methods that are strongly based on contextual information. This lays foundations for advanced data analysis and knowledge-driven machine learning.
- ScaDS.AI intensifies research on machine learning for dynamic and temporal graph data, which is an important but less explored research field building on strong expertise in the area of large scale graph analytics.
- Significant progress will be made on iterative development of AI methods and human interaction though conversational AI.
- To further improve AI foundations, ScaDS.AI performs research in novel neuro-inspired information processing methods that promise significant improvements of existing state of the art.
- The research on flexible data-driven AI workflows is intensified, in particular for large-scale machine learning on large computing infrastructures. In that context, methods for fast access to adequate compute environments for large-scale data analysis and machine learning are developed.
- ScaDS.AI intends to treat trust and explainability of machine learning and AI as a first class citizen and builds upon strong expertise in argumentation based systems. This also involves focusing on legal and ethical aspects and measures to prevent bias in machine learning.
- ScaDS.AI investigates the comprehensive preservation of privacy in machine learning to even allow analysis on person-related information while keeping rights of data owners.
In addition to these research areas, ScaDS.AI investigates challenges of applying AI in four application domains, which are security, sociology, hyperspectral imaging, and biomedical applications.
Figure 1 shows the three layers of research areas: (1) AI-Foundations, (2) Knowledge and (3) Applied AI. In addition to the layers, two columns represent the two crossing areas of research (Scalability and Society).
More details about the Big Data Research in ScaDS.AI can be found here.