Heeeh... I always find it entertaining and distracting to monitor and participate in heated arguments between proper musicians and ignorami. When however the argument is between educated, intelligent fellow composers I find it less so, especially when I believe that they really do agree on the salient aspects of a discussion.
So, with Zemlinski's String Quintet coming through my headphones, I'd like to submit my thoughts on this.
First, the short version:
There is no argument between the quality and artistry of a great soloist's or ensemble's rendition of a quality work, to that of the greatest programmer's with the best samples and technology on the planet. Were it not the case, Williams, Morricone, Shore, even HANS with his ultra-MECHANICAL, intentionally bereft of any emotive characteristic in his music, they would all hire programmers to record their soundtracks, instead of engaging orchestras to do so. Further, all great conductors would collaborate with the same programmers for that one perfect performance of each and every work in the classical canon.
Now, for the extended version:
Thirty years ago, those of us that were around -if pretty young- we all marvelled at the great achievements and quality orchestral simulations of the Synclavier, PPG, and Kurzweil. Some of us, more fortunate, also got their hands on some of these machines. Younger composers here can still investigate the quality of those (and other engines') outputs on YouTube, but it would be hard to imagine the awe we felt back then, our feeling of how close those technological marvels had come to perfectly simulate live instruments in an unprecedented way, because young people's ears have been spoiled by today's sample libraries' higher standards.
Similarly, thirty years from now, I anticipate an equal advancement in the quest for the Holy Grail of realism, and I can imagine composers in 2048 looking back at VSL's and other libraries' achievements during the last 10-15 years and appreciating those companies' accomplishments, but outright refusing to work professionally on today's platforms in favour of what they'd have at their disposal in their own time (how many of you still work with the Performance Tool? And don't tell me you didn't think it was a miraculous tool in its time). By then, there could be algorithms to perform a trombone line like Lindberg, or a flute line like Rampal etc. - like an evolved version of Logic Pro's Drummers feature. Who knows...
Orchestras on the other hand, they do mutate and evolve but in a different way. In fact, there are many music critics that prefer the Berlin and Vienna Philharmonics of old to today's. It's not worth getting into that huge discussion here, but what I am proposing is that those orchestras will sound - comparatiely to samples - pretty much the same thirty years from now, as they did thirty years ago (comparatively!).
It is my contention that composers of that time will still prefer, not just the NY Phils and the Concertgebouws, but any professional ensemble that will put the requisite amount of time and effort in the performance of their works, than their computer simulations, even with all the anticipated improvements thrown in.
I forget and am lazy to check who mentioned Pollini's playing having changed, and it is a great argument supporting the animistic quality of music compared to the other -static- arts. That is one of the qualities that separates 'Music' from them. A manuscript is only the DNA code of a work. It definitely prescribes an inevitable, but only general musical outcome, where the details are ever protean. Otherwise, it would only be necessary to record every symphonic work once with any competent conductor, or an instrumental work with any competent soloist.
Thus, a sample-simulated performance of a work provided by the composer (IF he is a competent programmer) can be considered a performance map of his work (maybe there is a PhD in this...), and of his particular vision and intentions, much like Stravinsky's own recordings are invaluable to a serious student of his music. I have never heard anyone however averring that those are the best recordings of that composer's works. Same with those of Rakhmaninov. They are a starting point and that's it. His performances are barely considered more authoritative than Hoffmann's, Richter's, Gilel's, or Horowitz's. And that is the limit of the artistic merit of a sampled performance at best, in my opinion.
I view samples as an invaluable tool for composition, as others view the arpeggiator and pattern generators. But mostly I view them as the Godsend of getting decent (or any) performances of my works. They are perfect in tempo and intonation, as they are inhuman and "dimensionless" when compared to a live ensemble (even a bad one).
They have to be so. A violin sustained sample for example MUST be perfectly bowed and perfectly neutral in execution and emotion, so as to be useful in the simulation of every and all symphonic works of the last 250 years, and all movements, fast or slow, serene or intense, requiring that articulation. A competent human's performance of the same note MUST be similarly perfectly bowed, but also must be richly endowed with context emanating from the work's meta-data if you will (composer, style, period), and structure (what came before, what comes after, import of phrase, meaning, spirit, etc.), and all this can effect the performance of that single note, and changes that will all occur within the performance of that one note. Even with all the 'Humanizing' features and articulation drawings in a DAW, I have yet to encounter the suppleness of human playing.
In conclusion, I don't see how there has to be a debate between orchestras vs. samples, as a) one doesn't preclude the existence and usefulness of the other, b) they each bring their own unique great attributes to the compositional table. Let us rejoice and give thanks for our access to at least one of them, always aspiring to even more and better access to the other (both actually).
Happy Easter (belated to the non-Orthodox).