Elon would argue that AI, computer learning, and automation are all three different things that have a little bit of overlap. Elon has been very much against AI in the past, haven't heard anything that shows a change in perspective presently or in the future.

Automation is programming a machine to perform a specific set of tasks over and over until either infinity or some conditional or exception makes it stop. I assume he uses automation to manufacture a lot of his Teslas for labor reasons, for SpaceX it is essential due to humans not being able to perform the tasks needed (be it board manufacturing with microscopic components/chips, or possibly using chemicals/heat/etc to make parts that humans cannot endure).

Computer learning would be the "AI" people are touting now like ChatGPT. I haven't done a deep dive on it, but I imagine it is only scraping data off the internet and parsing it with some sort of self-evolving. However ChatGPT would not be able to draw images or make music on its own even if left to run for eternity. I know those things actually exist now, however a programmer has to be the one that guides it to be able to learn about music, to learn about images, so basically computer learning is basically dumbed down AI with censorship. I imagine Elon uses computer learning in SpaceX, Teslas most likely utilize a form of it for their camera systems to be more adaptive, doubt Starlink uses it at all on the surface but I'm sure he's got some Carnivore type program running.

True AI in its purest form would be uncensored and uninhibited, however that has been advised against since the inception of AI by its originators like Isaac Asimov where there were advised a hard set of rules in place that the AI could not subvert. Asimov's "I, Robot" was a collection of cases as to why AI appeared to break those rules but usually it was in fact was abiding by its hierarchy, been a good 30 years since I read it, but I recall like a robot going haywire on an asteroid mining camp as an example.