I don't think there are many definitions of machine learning that claim the models to be intelligent. Most of them limit the term to models that can be built from data.
Learning is a skill that not necessarily comes with an "intelligent" label attached to it.
Have we even defined 'intelligent' might mean? As in, we had the Turing test as a bar and we are close to that already. What is intelligence then, last I checked, there wasn't a definitive answer do it. We'll need it so that we can label AI as I properly - or maybe we don't care so much... If it's close enough...
Once we start seeing cheaply made, imported yes/no engines (masquerading as AI or knowledge) flooding the market, the definition of intelligence will be lost on marketing anyways (unlimited data, superfood, etc)
A predictive model, whether created by ML (regression, SVM, NN, whatever) or something rules-based born out of data analysis is reliant on quality data, which can be expensive to get. There is also a catch 22 where most of the models that are easy to make aren't practically usable because they're not needed in the first place, like a model that tells you if it's a nice day outside; most people would probably take a look at the weather and decide for themselves. On the other hand, a model predicting optimal stocks to buy or self-driving car models are worth a massive amount, but are also really hard to make. Companies will obviously try to sell bad or cheaply made models and may be successful on a small or niche level, but I think most people will recognize the utility and efficacy of a model based on the difficulty of the task it accomplishes relative to their own ability in that task, regardless of buzzwords associated with it. However, a lot of powerful modeling libraries made by really smart people are open source, so maybe what I'm saying is moot apart from sourcing the data.
Learning is a skill that not necessarily comes with an "intelligent" label attached to it.