Hard disagree, a great deal of procedural art is pleasing and artistically worthwhile even if you don't consider the method it came to be - although usually it's fairly apparent.
Hard disagree with your disagree. Very few procedural artists seem to have a good feel for the medium they're working in, or know much about art/music history and current trends.
They either have a very shallow understanding ("I made an image" = "I made art"; "I played some samples in a browser or trained a GAN" = "I made music") or they have an academic misunderstanding, based on a decent technical understanding of the grammars of the medium, but limited insight into how creative originators can use grammars expressively.
There are exceptions, especially in visuals, so you have a point about Sturgeon's Law. But there are still far too many people treating generative media as a technical exercise, and too few using them for impressive original creativity.
They said (and say) the same thing about DJs who use automatic beat-matching and mp3s. Fascinatingly Berlin still clings onto vinyl (for now). It's fair to point out that the current users of the tools aren't seasoned/world-class producers/DJs, and don't have that level of training and practice or their musical ear though.
It's amazing that these tools exist, along with those that can operate them, so music can be generated via code, live, and it comes down to (your opinion on) whether or not these tools ever grow past the "research" phase that they're currently in right now. IMO the tools will continue to evolve until they are useful in the hands of "creative originators [that] can use grammars expressively".
There's a impassioned community building and refining the tools and tooling, and its my belief and opinion - we don't have to all agree - that the algorave tools will eventually get to a place that's useful for creatives, and into their hands. Then the music being generated becomes the attraction, instead of the coding tool being a gimmick of an attraction.
After all, what's a drum machine, but a clumsy, analog "for" loop?
I want to mount a small defense (softly? narrowly? I agree with most of your specifics) of the state of the art(s). Not looking for a fight. :)
Procedural art is both new, and it isn't.
There's already a reasonably rich (if still fairly new in the sweep of art history) tradition of procedural art within the broader category of conceptual art. Some simple, accessible examples are Raymond Queneau's Cent mille milliards de poèmes, Sol LeWitt's line drawings, and Yoko Ono's Grapefruit.
These aren't quite procedural in the sense you mean, but they are instructions designed to produce art when compiled and/or executed. They (along with many other examples) shatter traditional-art assumptions more profoundly and thoroughly (well, I think so...) than the cubists, dadaists, and abstract expressionists. They fully-realize conceptual art. If a natural reaction to Duchamp's readymades or a Pollock painting is an exasperated but I/my kid could do that, these works hand art-observers the blueprint and call their bluff.
They foreground the gaps between the concept/idea of a piece, our interpretation of the idea, any given instantiation of the idea, and our experience of any instantiation. They highlight the difference between the traditional roles of the artist-as-thinker, the artist-as-craftsman/technician, and observer-as-interpreter--at the same time as they invite (or even demand) all parties to cross and blur roles. They expose the absurdity of the cults of the Author and the Original Object in a way that foreshadows what digital re-/production has forced most creative spheres to reckon with.
With this as a lens, it's possible to expand the family tree. Composed sheet music (most purely if a band/orchestra/choir interprets it from the instructions without listening to how it's "supposed" to sound) smells like procedural art to me. Screenplays and stage-plays (which are, of course, scripts) have similar features. A decent slice of performance art--even if not openly published as one--can still be distilled to a repeatable set of instructions.
Maybe it's hair-splitting, but from this perspective I see procedural art as a fairly rich tradition full of artists with a good feel for their media and the broad arc of their traditions. From here, I think it's easier to see modern-computer-aided procedural/generative art as a somewhat natural outgrowth of existing traditions--intersected with new technologies that I don't think we've caught up with yet.
I'll artificially cut myself off here; I have too many thoughts on what the hurdles are to put a bow on. In short: it seems fair to say computer-aided procedural art(s) are immature and have a skewed technician:artist ratio, but I think it feels about right given the sheer amount of work+play+interplay that it takes to establish radically new creative practices. A significant fraction of computational-creatives are still hand-rolling their own toolchains. I'm not even sure the sun has risen.
While there are definitely exceptions, they are few and far between. The good ones do get recognition, so it may appear there is more than there actually is.
I am myself mostly familiar with visual generative and algorithmic art. I recently looked up the difference between these two and it's somewhat relevant:
Generative art is when an autonomous system that is not the artist takes part in or is used for the creation. Usually these are people with an art background. But very often, the art doesn't look that intriguing visually, because that is mostly carried by the concept of how it was made.
For instance. You can hook up a plant to an Arduino and some motors that control a brush. Now it's a painting plant. But the paintings will be rubbish. However, now you move the plant with sensors to one coast, and the motor control paintbrush setup 5000mi away on the other coast. Now there's an extra layer of depth to the piece, call it "non-local distributed agent control" or something. The painting will still be rubbish, but now there's something extra cool to the story/process, which is part of generative art.
This is why I very much dislike peopling calling their work "Untitled 2736" with no other explanation. If you have the above thing, but you only see the display output painting titled "Plant Study 663" then what is the audience supposed to think about it? No you put a little plaque next to it with the story of what you did, at least. I was in MoMa NYC years ago and it didn't do this, so I really only enjoyed the pieces that I recognized from being famous, or the ones that happened to look pretty. Other contemporary art museums that I visited did it much better and provided info, I had a much better time.
Point is, generative art is not always pretty or pleasing, because it doesn't specifically set out to be.
Algorithmic art seems to originate from the old school computer hobbyist scene, people using the computer to draw fractals, geometric shapes, etc. At first mostly because they were cool things you could make a computer do. Over time, this grew into more professional art as well.
Difference is, algorithmic art often sets out to be pretty. It's just that you use algorithms as brushes, preferably creating things that you couldn't come up with otherwise. I'm going to talk about these on the visual medium first because that's what I'm most familiar with.
Problem is, algorithmic art seems to have a bit of a soft divide between people with an art or design background that can somewhat code, and the computer coders with not too much design sensibility. The overlap is of course where the magic happens.
On the "art side", their work often looks very nice, but sometimes it lacks the algorithmic depth to take it to the next level, or come up with something original. What I see often is that a specific routine, say a cool wiggly way to draw a line (instead of a boring regular line), gets used over and over again in different works that use lines. It's a bit like you made a custom photoshop brush. What you sometimes get is an algorithmic work that is the combination of a bunch of "shallow" algorithms, but nothing really "new". It's difficult because you really need a solid math or computational science background.
On the "code side", you get all these toy examples of amazing algorithms. But their creations are picked to showcase as many modes of operation. And the colours will be white/black/blue/red. Or it will be rainbow puke because they used HSL/HSV too enthusiastically. The focus point of the work will be centered on the screen without much thought if it might look nicer any other way. Without too much background in traditional arts, they will not recognize a known path and how to execute it well (because history) or at least choose not to. Say what you made, just happens to remind a bit of Gustav Klimt. Maybe there were random colours and some happened to be shades of gold. Now you can choose: will you try and incorporate other elements of Klimt's style, or perhaps some Jugendstil aesthetics, to really nail it down. Or is that exactly not what you want and you recognize it so you can steer away from it.
Please note that all what I just wrote is very subjective to my understanding of generative and algorithmic art (corrections are very welcome). Also the "typical" examples I stated are of course caricatures to demonstrate a point. Most people will be a more moderate mixture of these.
Finally, about the musical examples in the article, I'm getting a similar vibe. I can't really say if it's the "code" or the "art" side doing this. What I'm hearing in especially the melodies is my own first experiences trying to write a random music generator. It's what you get if you pick notes randomly from a chord or a scale. Broadly, it doesn't take in enough context from itself and the rest of the music. The rhythmical aspect of melody, the importance of accent notes and phrase-ending notes, are usually ignored. Also larger structures like ABAC are constraints that actually add depth. But picking notes randomly from a chord or scale is a one-liner and thus great for live coding. Unfortunately the naive approach will always have this vibe of "the notes are not wrong, but they are very random" and different melodies generated will blend into each other because they are formed of the same mush.
Again, very very subjective opinions, and I'd love to hear how others see these things :)