Yes I agree synthesizing is the future and technology to simulate very complex sound from scratch has been here for quite some time. It's called physical modeling and it's widely used by likes of Korg and Yamaha in their professional musical instruments trying to emulate real thing. Problem is that so far it's only possible to emulate simple instruments like brass, reed instruments and stringed ones to some extent. Forget about piano, for example! Limiting factor is huge DSP processing power needed for this technique to sound anywhere near the sound of multi sampled instrument.
More, something like car engine is so complex to simulate by above mentioned technology, that's practically impossible to do it without huge external DSP processing help like Sharc DSP farms, etc. There are so many resonating cavities and moving parts in your average engine, that modelling all this would took up any CPU completely(probably 20 CPUs!), not to mention staggering complexity of a such mathematical model.
In the mean time the only reasonable and DSP cheap alternative is multisampling real engine and crossfading between samples. Yes, it's compromise ... but the results are much better at the moment with this technique.
Try to watch any Alfred Hitchcock movie first with sound, and with no sound afterwards. He was master of suspense, but watching his classics without sound(music) is like, well ...
Having REAL sound is as important for immersion factor, as having believable 3D enviroment.
And, btw, there is NO synthisizer in the world (except hugely DSP demanding physical modeling synths which sound believable) which would give you a better REAL representation of orchestra string section or piano than sample replay, for example.