Vox Day's Adolescent AI Musings

Theodore Beale, who writes under the pseudonym Vox Day and is one of the fathers of the alt-right, has no sympathy for the "performative AI critics" who are "protesting in vain":

Like it or not, AI is here. Everyone has surrendered on the illustrative and music fronts because a) the economics of illustration and the effective performance of the human illustrators are too horrifically bad to survive even halfway-decent AI production, and b) the average individual's ability to appreciate well-composed music and high-fidelity audio is virtually non-existent.

Strange locutions aside, this is a weird opening gambit. It's not really an argument per se and at no point does he come remotely close to defining his terms; he states a conclusion and offers some flimsy premises and simply moves on.

But what qualifies as "effective performance," for example? Or by what criteria do we judge that the effective performance of human illustrators is "horrifically bad"?  How do you evaluate the average individual's ability to appreciate well-composed music or high-fidelity audio?

It's also strange to me that neither premise has anything to do with the other. Illustration is being replaced by AI because the old regime is no longer commercially viable, while music is being replaced by AI because of the limitations of the "average" human listener (whose relationship to commerce is unexplored).

With the caveats that it's impossible to tell how much of the post was AI-assisted or -generated and that the post is immeasurably more successful if considered as parody, let's assume it does actually reflect Mr Beale's thinking on the issue.

He continues:

The success and widespread adoption of AI music was not only predictable but completely inevitable due to the the success of three predecessors: punk music, rap, and the MP3 format. All three are objectively inferior the current state of AI music ...

Consider this Suno 5.5 track, a new mix of The Long and Lonesome Skyway ... you don't have to like the song at all to observe that a) the musical artistry and lyrics are better than every single organic punk song ever recorded, b) that there is far more melody, musical complexity, and fewer derivative elements than in every single organic rap track ever recorded, and c) the 44.1 Khz CD-quality contains more information and sounds better than every single MP3 track ever sold.

These things aren't debatable or subjective. They are objective, observable, and easily provable facts. You would have to be comprehensively ignorant of the relevant subjects to even try to dispute them.

All of this is really weird. Mr Beale's initial claim was that AI music is inevitable because the average individual has no ability to "appreciate well-composed music or high-fidelity audio" but proceeds to praise AI music for its exceptional "musical artistry" and "complexity" - i.e., its composition - and audio quality. Why would well-composed music with high audio quality appeal to people who can't appreciate it? My guess is that he's smuggling in some kind of commercial claim, i.e., that most music consumers lack the refinement to discern human-made commercial slop and AI-generated slop, so human-made music - like human illustration - is no longer economically viable. But then, why is the so-called exceptional quality of AI-generated music of any interest?

Mr Beale also introduces three predecessors to AI music: punk, rap, and the MP3 format. It's unclear why he homes in on these but it doesn't seem facially that sensible. Punk, for example, is generally quite hostile to the commercial, political, and technocratic conditions that allow for the advance of AI. 

Perhaps he's smuggling in another idea: that punk and rap and the MP3 format - being "low" forms - have so degraded the discernment of the "average" listener that they no longer desire "quality." That's meeting Mr Beale more than halfway. But also it doesn't really cohere. Punk and rap both, in many forms, champion individuality and individual expression; AI "art" is about the obliteration of the individual. Which is to say that, for example, a listener can enjoy a low-fi punk record because its low-fi-ness is part of its expression of authenticity without that low-fi-ness inherently or inexorably degrading that listener's sensibilities. You can listen to a Thurston Moore song recorded into an answering machine and still appreciate a high-fidelity recording of a Scriabin performance.

Different qualities are proper to different modes of expression, a basic fact that Mr Beale repeatedly elides.

All of this is complicated by the fact that Mr Beale seems not to be a very sophisticated listener himself. He had a short career in professional music with the techno group Psykosonik, but he doesn't seem to have mastered the "relevant subjects" he implies he has.

Take "The Long and Lonesome Skyway." It's a pretty bog-standard (and quite corny) symphonic pop metal thing. It's deeply derivative of various genres of metal and alternative rock dating back at least to the early 80s. The melody is, to my ear, pretty uninspired; there's very little variety. Far from being complex, it's a basic 4-chord song. (Just because the lead guitar is arpeggiating doesn't mean it's complex). And there's almost no dynamic variety in the whole song.

But let's take Mr Beale's claims about it one by one to try understand why his analysis is so weird:

• "The musical artistry and lyrics are better than every single organic punk song ever recorded." Are we defining terms like artistry within a punk idiom or Platonically or some other way? Is artistry technical or spiritual, consequential, deontological? What makes lyrics good or bad, and how do we understand those values within a particular idiom?
  
  He dunks, for example, on the simple chorus of "Anarchy In the U.K." without interrogating why such simple choruses are important to punk, especially early punk. ("Nazi Punks F*ck Off," anyone?)
  
  There's no formal or cultural or historical analysis, no treatment of what the function of punk music is, and no real engagement with voice or perspective, with why The Sex Pistols don't sound musically or lyrically like The Ramones, even though they're both writing pretty simple punk songs around the same time.
  
  Here's a thought experiment, though: could AI ever write a line like "Don't know what I want but I know how to get it"? Simple figure though it is, its composition requires a subjective sense of irony, of hopelessness and despair, of rage and cheeky antagonism.
  
  But let's pretend that we can all intuit a basic paradigm for judging good lyrics from bad. Are the lyrics of "The Long and Lonesome Skyway" better than those of "Aimless Arrow" by Converge? Not to mention that the Converge song is much more complex, formally adventurous, dynamic, and in many ways less derivative. It's also the kind of music that only works when there are little imperfections, when you can hear guys in a recording space succeeding but struggling to manage these alien guitar figures and drum patterns. It is a work of human ingenuity and it could never be mistaken for something generated by AI.
  
  In any case, which song has better lyrics? I think the answer is pretty clear. Presumably Mr Beale disagrees. But it is really beyond the realm of what we can debate, as Mr Beale claims?
  
  And formally, what about "LAX"? "Something I Learned Today"? You can't seriously argue that "Skyway" displays more "artistry" than "New Noise," can you?
  
  Also, what is the real sense in comparing "Skyway" - not a punk song - to the tradition of punk music? It's a bit like saying "There's way more mayo on this BLT than you'll ever get in a plate of coq au vin." Sure, but I don't turn to coq au vin for a mayonnaise fix.
   
• This latter strangeness is even more apparent when he talks about rap. Yes, there's more musical complexity and melody in a your average alternative rock song than there is in your average rap song. They're different modes of expression. It would be technically correct to observe that there are fewer moving pictures in an Emily Dickinson poem than there are in Kangaroo Jack. What is the import of such an observation?
  
  But the idea that the lyrics of "Skyway" are more impressive than, for example, the work of Nas or Kendrick Lamar is insane to me. I guess it's the kind of subjective thing we have to debate about.
  
  
• "These things aren't debatable or subjective. They are objective, observable, and easily provable facts. You would have to be comprehensively ignorant of the relevant subjects to even try to dispute them." This is such preposterous basement-dwelling internet-bro posturing. Objective, observable, easily provable? By what criteria? What are these "relevant subjects"? Music theory, music history? Is there a PragerU video that can explain it to me? This is Mr Beale's favorite rhetorical game: barrage you with wild claims that barely hold together and then if you don't agree it's because you're simply not high IQ enough to keep up, you don't have the relevant expertise, etc. etc.

But then Mr Beale gives up the game:

It took me seven years to write A Sea of Skulls, a very good book with 104 reviews and an average rating of 4.7. It didn't take me seven days to write Out of the Shadows with AI assistance, which has outsold its predecessor by a factor of 4.2x over the first six months of this year, has 98 reviews, and an identical rating of 4.7.

So this isn't, and has never been, about what is good, true, beautiful, which is a foundational concern in Mr Beale's other work. This is about commerce, not aesthetics. It's about the sheer nihilism of productive capacity over quality, over humanity. 

Then he gestures at middle ground:

What people tend to forget is that AI isn't magic, anymore than than a typewriter or a word processor is magic. AI absolutely requires the spark of human imagination and creativity.

All fine and good, except that analogizing AI to typewriters and word processors is obviously wrong. Mr Beale produced a terrible, derivative song using only AI, and the generation of that song was only possible because AI as a model is premised on theft. It wasn't Mr Beale's "human imagination" that made the song possible. It was his human imagination that guided how he wrote the prompts, perhaps, but the song itself is the zombie corpse of millions of bits of data pulled from real creators without their consent. It's a photograph of a photograph of a photograph.

Beyond that, though, there's nothing in his post that indicates that Mr Beale is thinking about any of this clearly enough or with enough care to properly guide these AI systems in the spirit of "human imagination and creativity." Part of what's wonderful about human art is that none of us has all the answers. Nonetheless, in order to arrive at anything like answers, we must be asking the right questions. AI doesn't challenge you to ask the right questions. AI celebrates you for asking the wrong questions, as clearly happened here with Mr Beale.

If the beautiful is the true and the true is the beautiful, then truth must guide how we engage with beauty, just as beauty must guide how we engage with truth. If you have not truth, you cannot have beauty. Mr Beale seems happy to jettison the truth in favor of whatever conjured realities don't challenge his noxious, anti-human ideologies, which is I guess why he seems more or less happy to embrace insipid AI-generated parody hair metal as "beautiful."