Was eine KI niemals können wird, wird auch der Mensch niemals können

Abstract In this article, I will argue in favor of a broad concept of AI that is not restricted to a particular kind of technology. Arguments based on such a concept will stand the test of time and will not become obsolete as technology advances. Arguments concerning the limits and possibilities of technology based solely on steam engine technology, for example, would no longer be relevant today. The same holds for arguments concerning symbolic-logical AI – and will probably be similar in the future with arguments concerning current artificial neural networks. In a second step I develop a technology-independent argument to show that no system – no matter how advanced it may be – can ever fully understand itself. Furthermore, I will show that with a broad concept of AI – which also includes biological systems, for example – AI with human-level capabilities is, at least, possible. However, if artificial intelligence with human-level capabilities is possible, then the fundamental or absolute limitations of AI capabilities would also be fundamental limitations of human capabilities. In other words: what AI can never do, humans can’t do either.