Computerized reproduction of orchestral music gets better all the time, and whether it's good enough for some purpose depends on the purpose. I would say that the percentage of needs where a software-based automatic interpretation is "good enough", increases all the time, at the same time as the general public (i.e. audience) gets more and more used to all sorts of artificial sound.
Here's an example of Note Performer 3 by Wallander Instruments playing a Star Wars score notated in the Sibelius notation application. Good enough for ... something?
The main idea with Note Performer is that in combination with a score-writing program it interprets traditional musical notation into a performance, rather than relying on a "programmer" musician doing the translation and feeding it low-level geek-oriented MIDI data. Because Note Performer is driven by higher-level notation symbols like hairpins and slurs from a notation program like Sibelius, Finale or Dorico, it knows (or tries to reverse-engineer) a bigger picture. I'm not sure (after looking at the PDF manual) at which software layer the interpretation from symbols to data happens in each case, but anyway, Note Performer constructs a fluid continuous flow of expression, controlled by musical symbols. (And it needs a one-second delay for doing this)
With traditional sequencers and sample libraries, everything needs to be "programmed" in terms of, say, note velocity numbers, key switches, control changes and program changes, which end up triggering samples and switching or cross-fading between sampled notes. Wallander's synthesis engine is not based on entire sampled notes, but a "component-sampling + acoustic behaviour modeling" approach where each instrument's sound has been sampled as small components that are used to construct the whole sound in an additive fashion, guided by complicated behavioral instrument models which decide how much of which components are needed in order to produce the requested notes and pitches with the requested expression parameters. I'd say it's more of a synthesizer, definitely not a sampler. And it's not based on physical component modeling either.