The heart of that issue is that, broadly, AI is “something fine by a computer that would normally need to be done by a human.” When I was younger, the main (initial) goal of people developing AI was have a computer read a page of printed text out loud. Them, not too many years later, everyone had OCR and apps that could convert text to speech, so no one considered that AI anymore. Most ML applications would have been considered AI not very many years ago, but now they’re getting common enough that they didn’t meet the “needs to be done by a human” part now.
It’s hard to come up with a crisp definition of AI that doesn’t change over time.