Yes, DL is a sophisticated tool for the most intractable of tasks, and for most problems is like using the Death Star to crack a nut. This is well known and should just be a normal known thing any analyst of any flavour should have in mind - if you're using DL, especially anything not really big or that isn't natural language or image related, it should be for a good reason, because either a random forest, a GAM, or an autofit ARIMA will get you 80+% of the way there 80+% of the time for tabular data. Not everything needs to start with the biggest guns.

Perhaps suggesting that a lifestyle of doing nothing but waking up in your cookie cutter suburban house, driving 90 minutes to the same workplace, spending 9 hours there doing the same thing, driving 90 minutes home, and watching the same television programming for 50 years isn't part of the recipe for a good life?