I'm a CS professor. The problem with AI is that allowing it and disallowing it both lead to awkward outcomes.
Suppose you allow it. Then what are you going to ask students to do? Implement bubblesort? No. That would be pointless -- AI trivially generates all boilerplate code. Of course Chegg long ago broke the oldest and best coding assignments, which consisted of implementing classic algorithms from pseudocode. Still, all traditional undergrad assignments are out the window.
What *can* you ask students to implement? Realistically, they should be able to do just about anything. The order to "Recreate Facebook" is a valid 2 day HW assignment. But that's so broad that it's ungradable. And the students' ability to do the task means very little about their inherent ability.
So, suppose you *don't* allow the use of AI. Then almost everyone will use AI anyway. Now you're not a professor, you're a detective, and everyone in your course is a suspected criminal. If students are smart AI work is easily disguised. Then who are you giving A's to? Cheaters. Solution: Give everyone an A. Educational value for most students, who need to be threatened and cajoled to do work: Zero.
Does that mean I'm saying that AI makes everyone a genius and you can't tell the difference anyway? No. The pinch happens when you get out on the bleeding edge and try to do something truly novel. But that is not how instruction of any kind traditionally works. Things at the bleeding edge are incomprehensible to students. Asking students in CS 101 to blaze a new trail is a stupid assignment.
Conclusion: Things are very broken and many students are in trouble. The only thing you can really do to educate the typical person (who requires cajoling and threats) is to lock them in a Faraday cage for four years (and they would probably still cheat). On the other hand, for the *very* rare individual who is self motivated and just wants to learn, it is a golden age.
I suppose we should really be teaching students how to use AI to educate themselves.