Prof. Dr. Adrian Paschke is Professor of Semantic Data Intelligence at the Einstein Center Digital Future (ECDF) since mid-October 2019. The professorship is a joint appointment of Freie Universität Berlin and Fraunhofer FOKUS. Since 2016 he has been Director of the Data Analytics Center (DANA) at Fraunhofer FOKUS and since 2008 Professor and Head of the Corporate Semantic Web Group (AG-CSW) at the Institute of Computer Science, Department of Mathematics and Computer Science of Freie Universität Berlin.
In his research, Adrian Paschke deals with two artificial intelligence (AI) technologies: semantic AI and machine learning. "In my work, I combine both areas of symbolic and sub-symbolic AI. Because on the one hand it's about model-based and semantic knowledge and on the other hand it's about learning the computer from data," he says. As an example, he cites the semantic plausibilisation in image recognition. "When a cat is in a picture, we humans recognize it because we have learned what the animal looks like. For this we do not necessarily need many different cat images, but can learn from one and the same image, e.g. from a children's book. In order to make sure that a computer recognizes a cat correctly and does not confuse the hair of a human being with that of a cat, for example, the recognized image information can be semantically plausibilised with additional background knowledge, for example that in a crowd of people a cat on the head of a human being is not plausible and is therefore hair. Such additional knowledge from semantic knowledge bases, such as ontologies, can be used by the machine for interpretation and knowledge conclusions". explains Adrian Paschke.
These technologies are also important for autonomous driving, e.g. to make exceptions recognizable and to interpret situations correctly. "We need to train computers in such a way that cars, for example, can recognize people on the motorway and assess this as a potentially dangerous situation in order to make the right behavioral decisions in the sense of a normative- and value-based AI. However, there is often too little video material with such exceptional situations to train a machine," he says. Therefore, semantic models are needed to provide the computers with the necessary knowledge and simulation models with semantic knowledge to generate sufficient training data for machine learning.
In addition to researching semantic AI technologies and knowledge representation standards, his research deals with Edge Intelligence applications, e.g. in the field of Industrial IoT - the Internet of Things in Industry. "We are talking here about AI at the Edge. We move AI functionalities as close as possible to the machine that produces the data," says Adrian Paschke. If the data can be used directly in the Edge, for example, the automatic monitoring and maintenance of the machines can be improved. "Our goal is not to store the large amounts of data in a cloud, but to process them on the machine, i.e. in the Edge, with AI. This also increases the protection and privacy of the data," says Adrian Paschke.
Another exciting field for him is quantum-supported AI. "We can think of the first applications in this area, but it will take some time before we actually get there. It is therefore very exciting for me to work at the interface between basic research at the FU Berlin and early application with industrial partners at Fraunhofer FOKUS," says Adrian Paschke.
The ECDF is an interesting network for him. "I'm really looking forward to the exchange with my colleagues," he says. "As a semanticist, I am used to working interdisciplinary with domain experts from various fields in semantic knowledge modelling and to using my application-oriented AI research to benefit the transfer between science and industry". Especially in the areas of normative and value-based AI systems and digital curating for knowledge work, I see many opportunities for cooperation.