Those are interesting responses. I do not agree completely with polarbear, in that there are some things that sound completely real with samples now. For example that Beethoven string quartet by Bacal fooling a violin teacher or as another example a piece for organ or certain combinations of winds or percussion. We can easily fool people. But I am not interested in the "fooling" - it is more the use of these sounds purely as tools to create expressive performances, no matter what the original is.
I agree on vibrato's point about using the samples in non-traditional ways. I often think that Ligetti was trying to use samples before samples existed. His orchestral pieces, even some of the vocal pieces, use sustained or extreme range sounds that are very difficult and non-characteristic playing for a live orchestra, but are perfect for samples especially in how they emphasize timbre. If you write 50 bars of a sustained low e for live basses they will get very angry with you, but if you do it in your sequencer and sampler it is a piece of cake. Or a series of pedal tones in a bass trombone, etc. etc.
I do think that the sampling of more and more special or even completely new playing styles should be a priority for the most advanced sample library. In other words, to provide more and more expressive range to the sound that is available and to create new avenues of sound that may not even exist in avant garde music.