Artificial Intelligence has nothing to do with the capabilities, and everything to do with the approach. If I hand-crafted an algorithm that could reliably spot stop signs in pictures by applying a mountain of heuristics that I came up with, AI would not be an appropriate label. If I instead made a program that could detect stop signs after being fed thousands of pictures of stop signs, then I'm comfortable calling that machine learning, or AI, because the approach was based off modelling intelligence, i.e., the ability to acquire and apply knowledge and skills.
The former was the original use for "AI". The latter was called "machine learning" largely to contrast it with "AI" whose reputation had tanked due to hype.
