I learned about simulated annealing, gradient descent, perceptron feedback methods in the 1990s, when those were still being called AI.
I was the pm for multiple teams that used ML classifiers at a major Internet company to get serious work done, for a decade. Nobody called it AI.
From where I'm sitting, the frenzied use of the term AI on a grand scale for fundraising coincided pretty much exactly with general purpose transformers applied to language models (and to a lesser extent diffusion models).
I feel like calling all machine learning AI is confusing, because it confuses actually well-designed systems that do real stuff with an emperor-has-no-clothes bullshit mania.
It feels like maybe you're trying to extend the AI halo to non-llm, non-gpt algorithms because you think it will improve the esteem in which the latter type of system is held.
I think the AI branding is a stain, I think there is going to be justified and ferocious backlash, and I would want to keep a perceptual moat between "AI" and whatever I'm building, even if at some point I do want to write code for an npu.