As often with Wikipedia articles, It is the academic view that dominates. It may historically accurate, but it is still misleading.
If you are not familiar with how music is produced today, I can tell you there is nothing "high art" about this anymore. Gazillions software and plugins are designed on those premises. And this not only for Algorithmic/AI uses, or even just in electronic music genres.
I’m simultaneously in agreement but also drawn to defending the idea. I do think a real distinction can be drawn between human and/or physical chance as part of music composition, and software random numbers. Software random numbers are extra boring and predictable - you know everything about a distribution and it’s behavior a-priori; rolling electronic dice doesn’t add anything interesting. Some (not all, but some) of the examples in the Wikipedia article are not synonymous with software randomness. Chance in a music composition (or any work) can be intentional and arbitrary without being random, in a way that software does not and cannot really capture, at least not yet.
There is also a continuing academic following of the work of Ives and others in music composition circles, and I feel pretty sure most of them would disagree vehemently with the assertion that software plugins even remotely represent the same concept as what they’re trying to do.
But maybe to agree with you somewhat, there has for a while been a weird fascination with randomness among “generative” artists, which might be tied up in popular narratives about free will. Now that computers have gotten really powerful, it’s become more clear that randomness might add salt here and there, but relying on it makes art overall muddy and lacking in intention. I think people are generally coming around to find that randomness doesn’t really add surprise. Same goes for AI at this point - each new neural network is magic for a short while, but then everyone starts to notice it’s producing the same thing.
I believe the digital randomness can be "tuned" to be very musical (even though it often is not). Sort of like how some modern video games with random elements have a "fun random" rather than mathematically accurate random generation. For example, League of Legends "crit smoothing".
The digital modules by 'Make Noise' are a favorite of mine for this, especially the collaborations with 'soundhack'. They can't listen to the other instrumentation in a piece, but I'm positive they have some built in idea of holding back and then letting loose. They take a lot of inspiration from Don Buchla's analog random generators which were meant to be a semi-controlled random source rather than having a white-noise based random core.
'Mutable Instruments' actually has some modules that can take a rhythmic trigger sequence to generate appropriately matching random fluctuations.
I do agree most plugins and digital hardware with random are rather soulless PRNG's. But it's possible to make something that feels like it has a personality.
> a real distinction can be drawn between human and/or physical chance as part of music composition, and software random numbers
That's a good point.
> weird fascination with randomness among “generative” artists, which might be tied up in popular narratives about free will
A lot of modern art has a more "didactic" than aesthetic function. Once we assimilate/feel those (intellectual) lessons, they aren't of much use anymore. But the weight of habit/legacy/prestige lives on.
> each new neural network is magic for a short while, but then everyone starts to notice it’s producing the same thing
I don't agree. Deep learning is indeed lifeless/unconscious. It is also more akin to a complex system [1] than a random one. The way such systems create order out of randomness is fascinating to me. We don't know what life is (yet). So, we really can't say what a "lifeless" complex system is capable of.
If you are not familiar with how music is produced today, I can tell you there is nothing "high art" about this anymore. Gazillions software and plugins are designed on those premises. And this not only for Algorithmic/AI uses, or even just in electronic music genres.