Self-learning robots and AI systems can provide support in space, in the deep sea or in the event of disasters - by exploring unexplored terrain or carrying out work in dangerous regions. To do this, the technical systems must act as autonomously as possible, but at the same time always remain under the ultimate control of humans. Jürgen Beyerer explains in this interview how these requirements can be implemented and where there is still a need for research. He is a professor of computer science at the Karlsruhe Institute of Technology (KIT) and director of the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB. At Plattform Lernende Systeme, he heads the "Hostile-to-life Environments” working group.