The thoughts presented here start from Gregory Bateson’s logical categories of learning, in order to be able to better classify machine learning on the one hand, and on the other hand to ask which conditions a more general concept of learning should satisfy, which, for example, also includes the learning of slime moulds.
