Nobody's proposing "just a few years". Both Musk and Bezos have grand visions that require difficult, complex technical and financial problems to be solved.

First up is lowering the cost of escape velocity such that "casual" flights are financially feasible. This includes reusable components, a very low failure rate, and an ecosystem of employees, engineers, and culture that supports it.

If M/B achieves their stated goals in this domain, then the next steps can be practically executed.

A big part of astronaut training is knowing how to fly the craft and knowing what to do if something breaks. If the machine is automatically controlled and reliable, you don't need as much training. (Maybe they'll send a trained person or two along with the tourists?)

Not sure I follow that logic. More automation might mean you won't have to use your training as much, but how does it mean less training? In the event all the automation fails you still have to know everything an astronaut would need to know were there no automation.

For Starship, there's absolutely no way a human could fly it anyway. Even if it weren't fly-by-wire, a human could not control the drag fins and the engines precisely enough to land safely. Even on Crew Dragon, which will (hopefully) fly in the near future with crew onboard, there isn't a hardware joystick. If the vehicle fails to the point where the computers aren't working, your only option is to deorbit. These spacecraft weren't intended to be flown manually.