The aim of the project "Organic Machine Learning" (OML) is to break up the conventional, rigid approach to training and deployment of machine learning systems and to develop machine learning methods that are similar to organic learning, especially human learning, in which systems learn throughout their lifetime - especially during deployment.

The way of learning should become more organic. Instead of learning on very large, clean, and well-structured training data that had to be elaborately prepared, the systems developed in OML should learn from heterogeneous, less or not at all prepared data, as it is the case in the world, and be able to get by with less training data - similar to the human being. Different sources - for example, interaction with people and personal experience - are to be combined multimodally and learning is to be focused more specifically on cases of uncertainty. To achieve this, the systems must be able to recognize in which cases they are uncertain or where further learning is necessary. Moreover, learning systems should no longer be just a "black box" whose function outsiders have no insight into. Instead, they should be able to explain their decision making, their actions. The ability to justify their decisions makes the systems' decisions more acceptable to humans and their use for many applications in the real world possible.

Finally, the developed systems are to be brought together in a robotic overall system. In a scenario of interactive robot programming, the robot is to learn new skills from scratch by learning from physical, visual and verbal interaction with humans as well as from its own experience - similar to how an apprentice is taught by his master.