There are a few definitions:

AI in a slang or sci fi common lingo kind of way is any tech that appears more human. (At one point in time the fill tool in Paint was called AI.)

AI in a technical definition is any algorithm that makes a series of guesses to get a good enough answer, because finding the perfect answer would take too long to compute. Therefore, the study of AI, like if you got a degree in it, is many ways of study on how to make a series of educated guesses in the most efficient way possible and then putting that into code. (Source, I have an AI degree before ML became popular.)

ML is called AI because at its heart it's making a series of educated guesses. Say you have a million photos of animals, the first photo you show it is a cat, you ask the ML, "What animal is this?" It doesn't know, it's the first time it's seen an animal, so it guesses wrong. You give it a second picture of a dog, "What animal is this?" and so on. By the time it's seen the 100 thousandth image it has a pretty good idea what a cat and a dog looks like and when asked it guesses to a high level of confidence.

ML is not classical AI, where you write an algorithm that makes educated guesses, but an automated form of AI. It's called AI because 1) It looks more human than previous generations. and 2) It's making a series of educated guesses as efficiently as we know how to.