Introducing AI into education will be non trivial. This is difficulty is observable already in my experience. When offering advice on how to learn how to progr

Introducing AI into education will be non trivial. This is difficulty is observable already in my experience. When offering advice on how to learn how to program, I have to heavily recommend that students try their best to avoid the use of AI. Whereas previously the best advice I could give was to "build something", it is now possible to build a piece of software without understanding it at all. I have observed this myself with Rust; I have built a few programs now by repeatedly prompting AI models. I have even been quite engaged in designing the architecture, guiding the programs towards patterns that my intuition as a programmer says will be good for Rust too. The software works, but I can't help but feel I have learned nothing at all. Building something is now insufficient to learn, at least in the domain of programming. I feel there will be far more compilations we will have to address in order to benefit from AI in education. That said, I am still optimistic that it will be a net positive force.