I think we are still a bit of ways away from AI taking over debugging. From my experience playing with GPT it quickly starts failing if the questions get as complex as debugging an intricate problem.

However, if this can take some effort away from me having to do manual test plans, I'm onboard.