I'd like to read your opinions about having a Master and or a PhD to develop cool machine learning projects. Do you think I could achieve great things without them? What advantages do you believe having those degrees have?

I study in México, so debt wouldn't be an issue, the time is what concerns me. Also I tend to change of interest rapidly, so idk if self study would be a better option for me.

Depends, getting one implies you'll dedicate yourself to it, which i doubt many people could do if they self taught.

That being said, rarely have I seen people in this industry with masters or phD.

My advice is to have an end goal (simple project) and just work towards it, through that you'll learn all you need through actually doing.

@D--M btw. Do people you've meet with those degrees earned more than regular engineers?

You can achieve great things without them. But it might be worth it to stay not because of the degrees but if you see some interesting research project to participate in. Various grants can help ease the money loss. And compared to enterprise you have much more freedom for experiments. In enterprise getting results matter, in research, even mapping dead ends counts.