I'm not diving into hairsplitting here - there's an academical field that researched AI, and as part of that field, machine learning is investigated. That doesn't mean there's anything intelligent about the core mechanism of LLMs. The part that has many people confusing it for "intelligence" is that with actual - deterministic - algorithms, the in and output are filtered to and from tokens, according to grammar rules of languages, to avoid the most glaringly obvious garbled output.