It’s not like using AI is some elusive skill that needs a while lot of training, the whole point is that it in and of itself responds to natural language and responds in kind.

Now building AI systems, training, prompt stuffing, etc, those do warrant some education, but using them… The biggest thing to teach is that their apparent confidence is untethered from their accuracy, but that hardly calls for a significant role in education.

With progressing to calculators, you are offloading basic fundamental tedium to chase larger complexities. Biggest issue there is what complexities are you chasing using the LLM that the LLM can’t do in an education context? Keep in mind that for evaluation in education, we frankly don’t push things into the unknown, because the evaluator has to know that it is solvable, how it is solvable, and the solutions should be viable as a basis of comparison between students. Pretty much the moment it becomes fodder for education or job interviews, LLMs are going to be pretty good at just doing the whole thing. It’s still moderately useless for my professional area, but it can largely whip up a slack knockoff with no problem. A colleague told to evaluate LLM “right now” resorted to “interview ware” for lack of better ideas and it knocked it out of the park and got him excited, then when he tried to put it to work on less well trodden work and was shocked that it had deteriorated to totally useless when he actually needed it.

In short, acknowledge the models and platforms and their usages, but I dont think they should factor significantly for students or job interviews.