“Modern Artificial Intelligence” is a branch of Computer Science which triggered a hype in CS research and education and which is associated to key words as big data and deep learning. It is worth noting that “traditional Artificial Intelligence” is - with the exception of neural networks - conceptionally rooted in Logic: Expert Systems, Logic Programming, SAT solving, etc. are topics which could not be taught without Logic as fundamental discipline. One has, however, to admit, that the foundational discipline for modern AI is Statistics, rather than Logic. We will discuss three challenges for modern AI, which give evidence that Logic remains an important topic for Students in this area.

1. Notion of computability. Currently, there is no theory available which would allow to reason about the scope and limits of the computational power of modern AI. Theorems like those of Church and Turing are fundamental for the understanding of deterministic computation, including the important area of computational complexity. Apparently, the statistical foundation is not suitable to provide such a theory for modern AI, and for any progress in this direction, Logic should be crucial.

2. Hybrid AI. Despite all the success of modern AI, as in image recognition, translation, medical applications, and other areas, a substantial extension of the field of applications will require a combination of it with classical AI, i.e., the possibility to recurse to logical reasoning.

3. Causality. The problem of causality—or better: the lack of tools for the analysis of causality—is widely discussed in modern AI. Although even Logic still struggles with a compelling theory of causality, it is evident that any future solution will involve logical components, even if it would only be for crosschecking.

Thus, a CS curriculum for AI students which neglects Logic, bears the risk to disconnect these students from developments essential for “next generation AI”.