If machine learning was packaged and sold as "applied statistics," most undergraduates would think it's a boring as **** topic of study. Yet, that's exactly what it is. A "machine learning scientist" is more or less a computational statistician. A "machine learning engineer" is more or less a data engineer who understands statistics. The term "machine learning" is just a form of branding, as the word "learning" implies *intelligence*, which computers presently *do not* have. That said, it's disingenuous to equate AI with machine learning. This is because AI is really more about the *application* than the *method*. Cutting edge natural language processing is currently done via statistical models. But natural language processing is so much more than statistics. Robotics is a combination of control theory & computer vision, both of which are built on top of statistical models; but that doesn't stop it from being genuinely "cool." The trouble with machine learning - or applied statistics as I prefer to think of it - in industry is that it's typically employed for boring problems with boring solutions, like targeted advertisement or retail analytics. Don't blame the method - blame the application.
https://reddit.com/r/cscareerquestions/comments/j7gimm/unpopular_opinion_actual_machine_learning_work_is/g84sfsq?context=3