Thanks for your response.
On the idea that theory is not as important, I do agree that this is certainly true. For most ML engineer jobs, you will not really need to understand a lot of the conceptual theory. However, a concern I have developed is that if so many new developers see this option, they will gravitate towards it. This means that the long-term progress of AI will be slowed, as we will have less people focused on research and producing the top papers on ArXiv. However, the degree of importance of this concern is debatable.
I definitely agree with your second point regarding not understanding the data and how to properly engineer features from it. This is a great point and something that I did not think of. I also agree with the third concern about validation and test data. I actually have made this mistake early in my research career.
Thanks for your ideas.