Thanks for your thoughts! They mostly match mine. Your first point around the people using the term scientist loosely is interesting. I have no doubt that causes tension in the academic community. As to what "scientists" meant in the past, I would take it one step further; scientists usually referred to someone with deep knowledge of the natural sciences, which theoretically would not include PhDs in the computer science or statistics space. I think that distinction is not as strong as it used to be; otherwise, no one would be a scientist in the ML space.

In my mind (rooted on the natural science front), scientists usually fit in one of two broad categories; experimentalists, who perform experiments to test a hypothesis and theoreticians, who develop models to explain existing data to predict new results. If you look at that criterion alone, you could argue that some, not all, data scientists carry out that type of work daily. Beyond that, you mentioned an additional important component - the means of advancing the field by discovering new things and creating breakthroughs. I believe very few individuals with data scientist titles do that. We agree that "scientist" is a misnomer given this additional information.

There is only one small thing we may disagree on (if I'm reading your thoughts correctly), and that is the general signal you've given that someone must hold a Ph.D. in the field to be considered an expert in the space and contribute to research. Machine Learning is a new field, and as such, most of the main contributors to the area do not have that background. It was one of the reasons why I became interested in the subject many years ago; the diversity amongst individuals was a refreshing change from the stodgy professors I had in Physics who gated research behind a specific degree.