I see people refer to AI as "training wheels" but even training wheels force you to learn balance while riding a bike. They just give you more flexibility in how much you can lean. But AI, it is easy to just accept the answers and treat it like a black box. There is no need to practice the actual skills of a developer: designing algorithms, analyzing, testing (designing tests!), reading code deep with a code base and learning how it connects, interpreting asks and determining appropriate ways to solve those problems (a big part of experience is having fewer unknown unknowns. Known unknowns are incredibly valuable)
While all that is possible with AI, I don't see it happen often in practice due to other pressures. And with AI it seems there's just become increasing pressures to get more done faster. But a junior still needs time to learn. We all do, but juniors need more and have less experience and ability to find this time