Execution Environments & Innovative Computing Methods

How can artificial intelligence be optimised for specific systems? How do we secure devices that use AI against cyber attacks? How do we integrate AI into existing processes and devices? Our scientists in the Execution Environments and Innovative Computing Methods department are working on these and similar questions.

In order to embed artificial intelligence into our technology in a secure and uncompromised way, the entire system on which the AI is executed must be considered. To this end, the department is working on implementing and securing AI models at various levels close to the hardware. We are investigating the suitability of quantum computers for AI applications and optimising AI for high-performance computers and distributed systems. Securing virtual platforms for the permanent integration of AI software modules in vehicle technology is the subject of research in the Gaia-X 4 Future Mobility project.

We are also active in highly specialised application areas, such as the optimisation of AI methods in autonomous mobility and logistics facilitation, which require exceptional levels of reliability, but also adaptation to resource-poor environments. In distributed system environments, such as cloud services or decentralised data and service ecosystems, it is important that AI methods are provided in a device-independent manner. This is the only way to ensure that AI can be executed securely and efficiently on all participating devices. Our application-independent architectural approaches make it easy to transfer our methods to new tasks.

Quantum computing and other innovative computing approaches are an important aspect of the design of future generations of AI. We analyse the possibilities and consequences of new computing methods for artificial intelligence. In particular, the question arises as to how safe and comprehensible AI can be realised on quantum computers. The potential of the quantum world for the development of new methods beyond the improvement of classical algorithms will also be analysed. We also investigate how these new approaches can be integrated into existing classical processes and structures, for example in the context of cyber security in open data and service ecosystems and AI. Another important aspect is the question of what future attack scenarios on our IT infrastructures might look like and what defence mechanisms are needed when unconventional hardware such as quantum computers are used.