The issue is the overloading of the word “AI”.
“Machine Learning using Neural Networks” is a technique that can come up with decent but rough solutions to problems where it’s hard to come up with any solution.
“Large Language Models” is the application of “Machine Learning using Neural Networks” to natural language processing, and it is incredibly good at that.
The problem comes when people apply models trained for natural language processing onto other random problems just because you can formulate anything as a natural language problem.
As a general rule the more you spend up front, the less you will spend (in time and money) to fix and maintain the thing.
3D printers are finicky which is why they often become a whole hobby on their own.
As part of that, I’d strongly recommend you stick to one of the easier to work with materials (PLA and TPU seem to be popular rn). Those are good enough 99% of the time, and printing more exotic materials is more work. If you really need a better material, prototype in PLA and then buy a professionally printed final piece (I’ve personally used Shapeways a couple times. I wouldn’t call it cheap but for small parts it’s reasonable and the quality of the end result is quite good).
I personally am using an EnderV3 right now. It’s very customizable, and was one of the cheapest options when I bought it, but it tends to take a lot of debugging every time I want to make something.