Tacit knowledge—an impediment for AI?

The fact that much of our knowledge is tacit or implicit and must always remain so is often claimed to be a great obstacle in the development of AI. In a discussion of this claim, two different theses are distinguished: a weak thesis that many of our actions depend on tacit knowledge difficult to make explicit, and a strong thesis that all our knowledge has an ingredient of implicit knowledge which is not only difficult but, in principle, impossible to make explicit. Contrary to what may be expected, it is argued that the strong thesis is not a problem for AI, while some of the cases in support of the weak thesis do constitute such a problem. To wit, even in a programming language there are primitive operations whose meaning we know only implicitly, but that does not stop them from operating.