In addition to StuartC's comment, I would also like to point out that there is nothing new/revolutionary about this. I first worked on AFCS systems back in 1979 on the Lynx. There does seem to be an attitude, with some people, that making a helicopter easy to fly is somehow cheating. Yet I don't recall anyone from my aircraft-engineering days suggesting that our purpose in life was to make it more difficult for pilot's to fly the things.
One other aspect of RL flying, that I think is overlooked in FG, is that real pilots are not expected to figure things out on their own by trial and error. Someone who has done it before normally goes with them on the early flights. This QHI (Qualified Helicopter Instructor) would quite happily take over the controls to stop anything disastrous happening. Yaw-control and collective-control (power) are the two obvious examples where a QHI would only let a student take control gradually. So what's wrong with having a kind of AI-QHI for people who want to learn how to fly a helicopter? You can always kick him out/turn him off when you no longer need him.
Even a bit of yaw+collective damping can make an FG helicopter a lot easier to control, for example:
I believe the Bo105 also had AFCS (and it's from about the same era as the Lynx and also had a rigid rotor head). I think that the Bo105 is possibly one of the worst examples for a beginner because it requires about 45% back cyclic to lift off vertically /hover. This 45% is not too hard to maintain with a mouse, but it certainly is with a joystick. The EC145, for comparison, only needs about 5% forward stick during lift off/hover. Maybe the difference from StuartC's CH53 is the rigging of the main rotor head, i.e.: with no cyclic input, the lift is vertical. That's not cheating - just engineering.