In general, my opinion regarding machine learning and, in fact, also other tasks, is that Prolog is no less suitable to implement or teach these approaches than

In general, my opinion regarding machine learning and, in fact, also other tasks, is that Prolog is no less suitable to implement or teach these approaches than - for example - Python, which is currently often used in these areas. Dedicated libraries for machine learning are available and can be written for Prolog as well as for any other language, and Prolog of course also lets you easily interact with other programs via pipes and sockets. So, yes, I agree: Prolog may well be suitable for this! In practice, for statistics-based methods, the core logic will likely be carried out by dedicated hardware which you may have to program in its own language. Prolog may still play a part also with such approaches, for example to generate test cases and correctness certificates, and as a rule-based safety measure to prevent certain situations from happening.