They can do something that looks a lot like refactoring but they suck extremely hard at it, if it's of any considerable size at all.
Which is just moving the goalposts, considering that we started at "AI will never..."
You can't win an argument with people who don't care if they're wrong, and someone who begins a sentence that way falls into that category.
The guy who said "AI will never" is obviously wrong. So is the guy who replied that they already can. I'm not moving the goalposts to point out that this is also wrong.