An important question: Which do you want to do for 40 years of career time? I wouldn't want to do RF until Tuesday much less for a career. But that's just me because the RF engineers seem to be happy. It might even be a niche kind of thing. Maybe it pays particularly well.
When given a choice (seldom), I took digital courses.
It seems to me that the new emphasis is on AI and massively parallel computing. Literally everything will be driven by an AI over the next 20 years or so. But what undergrad degree heads in that direction? Applied Mathematics, EE (digital ?), CS, Business Admin? They all wind up there sooner or later albeit from different perspectives.
I am truly fascinated by the work being done by Google and NVIDIA. I suppose I should be jazzed by the work at Tesla but it seems their cars are attracted to police cars. There just has to be a story in there somewhere.
Side issue: The day won't come where I turn over driving to an AI. Isn't going to happen!
Just think about a graphics card with 8192 CUDA units (floating point number crunchers). The card is available today (if you can find one). Check the Specifications here:
https://www.pny.com/nvidia-rtx-a5000 That's a lot of Tflops for a beginner in AI.
https://www.amazon.com/PNY-NVIDIA-A5000-Graphic-Card/dp/B09G1Y6ZGTThe CDC 6400 that got us to the Moon was good for 2 Mflops and that graphics card is good for 70 Tflops - 35 MILLION times faster. It has been an amazing 60 years! Note: you probably can't write an OS that runs on CUDA units so there would need to be an Intel I9 <super whiz> (or multiple Xeon Platinum processors) running the OS and shoveling data in and out as required. Your job as a magician is to keep data inside the graphics card so that upload and download is minimized.
If I were in school today, I would be working on any aspect of AI.