Re: 50 years from now: what amp will guitarists be playing through?
I used to think that digital modeling would be the future (don't get me wrong, it has it's place), but I think ultimately tube-based amps will still be where it's at. The reasons I think this are two-fold:
First, there is an inherent delay in signal processing that is a function of the linear digital algorithms rather than the hardware per se. For example, an 80 Hz tone (low-E) has a period of 12.5 msec. There is going to be a 12.5 msec delay before a Fourier transform can extract the frequency for subsequent processing.
Second, tubes can do much more complex and interesting calculations that would need to be either done with a 5+ dimensional filtering map (e.g. clipping depends on frequency and amplitude, as well as a host of other things, including being dependent on what happened the split second before, as in 'sag') or you would need to simulate all of the analog parts in the signal chain. Some of these things simply can't be done in real time, but are easier to do with tubes. My analogy is with a pickup simulator that uses the input from a single bridge pickup ala a Line-6 Variax: it will never be able to simulate the complex mechanical interactions that occur due to phase differences that occur due to different fretting locations interacting with coil placement in a multi-coil guitar (either two singles in the notch position, or a humbucker's dual coils).
In essence, the timbre you can get from a tube amp is potentially much more rich and interesting than can be digitally modeled in real-time. It would be best to use tubes for what tubes do best, and digital for what digital does best.
What we might ultimately see it is a hybrid, where the amp uses tubes, but the circuit is digitally reconfigurable on the fly. For example, you could stomp on a footswitch and the signal path would change from a Plexi to an AC30 (assuming you have EL84 and EL34 tubes installed).