The fact is a modern turbojet, multi-million dollar flight simulator is several levels above the common computer program. The limitations are several: speed and memory of the computer; the completeness and accuracy of the data used to program and validate the computer; the size and placement of controls, panels, instruments, switches, gages, etc.; whether or not there is control “feed back” that is something other than generic or provided by springs or bungee cords, and the list can go on for quite some time.
However, there is little doubt that someone can get a good photograph of the cockpit of any airplane and use a computer to develop and present a reasonably good representation of that instrument – and this can be done for the entire cockpit – eventually. Now, the aerodynamic math model that is used can probably be either developed by a smart aerodynamics student in college or purchased rather cheaply from several sources. It is how you modify that basic model that will allow the “simulation” to perform anything like the a real airplane. Obtaining the information about what to modify and how to modify whatever it is can also come from several sources. For example, there have been line pilots who have been approached to copy down (or video) observed instrument responses in certain situations and provide that information to be used to modify the math model. The problems come when a given system on the “simulation” is asked to operate in conjunction with another simulated system – or even asked to operate at its limits. And things get really interesting when all the simulated systems are asked to operate in an integrated manner. Of course aerodynamic things like side-slip (and other non-linearities), slow speed and high speed operation, all of the stability and damping factors are completely or almost completely missing, control displacement and control feedback are completely subjective in all axes, and again, the list goes on and on. And, there are issues of sound, visual systems, and motion cueing that all come into play.
However, having said all that – there are some good things that can be derived from the use of a relatively inexpensive computer program that reasonably represents the systems operations of an airplane (this usually costs substantially more than MSFS or similar, however) – and if such a system is used in a spatially correct environment – checklist useage, cockpit flows, systems knowledge, and to whatever degree is allowed by the sophistication of the programming – this may include systems interoperability; crew interactions, and the like, may be accomplished quite respectably …again, dependant on the degree of accuracy involved and the limitations on the use of the device.
The whole idea of “simulation” is that it’s not real – it’s “simulated.” How well it is simulated becomes the question of the hour … and following closely behind is the question, how good does the simulation have to be to get acceptable training? Without trying to be overly pragmatic, the answer is dependant on the realism of the goals involved. Currently ICAO is considering a submission from the UK’s Royal Aeronautical Society that would, if adopted, provide a listing of what simulation features can be used to train pilots – depending on the pilot license level, the type of training involved, whether or not the pilot will be going to a more advanced simulator or to the aircraft for further training, and so forth. Simulation is here and its going to stay – what is needed is an acceptance of what is needed and what can provide that need.
MSFS can provide some things – but, as has been pointed out, a competent pilot will likely have more difficulty, initially at least, in trying to “fly” that particular computer “game,” and will have difficulty until he or she learns to adapt their existing skills (and that would be to “downgrade” those skills) to a degree that will allow them to successfully "fly" the game – and, personally, I can’t imagine why anyone would want to do that.