AI Engineering

In this research field, we integrate methods and technologies of artificial intelligence with engineering procedure models - in such a way that engineers are also able to responsibly implement AI-based system and solution modules in security-critical applications. The primary focus is on models and tools that serve the construction of security-critical systems or solutions, such as those required in the areas of critical infrastructures, transport, energy and aerospace. For these processes, we develop reliable evaluation and test procedures which enable the stability and controllability of AI-based systems throughout their entire life cycle. Through our work, we aim to increase the resilience and cost-effectiveness of AI, facilitate the use of AI in industry, and broaden the range of applications.

A particular aspect of AI engineering is the interaction between humans and AI systems: human-in-the-loop. At the Institute, we are explicitly dedicated to this human-centred AI and, for example, to the question as to how humans can contribute to the security aspects of both weak and strong AI. Humans can, for example, be “brought into the loop” when training machines, in order to enable decisions to be made concerning the exact form of an AI or to introduce new information that indirectly influences the training or supports system-based learning. The integration of human expertise also ensures that the AI not only incorporates this data, but is also connectable, i.e. it also takes human requirements into account.