Ethan Hein on how AI is harmful to learning (specifically musical learning) and I am reminded that Augustine of Hippo said that our ability to reason and our capacity for memory are essential to musical art

https://www.ethanhein.com/wp/2026/ai-is-harmful-for-learning-which-is-why-im-banning-it-in-my-classes/

…

Academic integrity is not my main motivation for the AI ban. It’s a factor, but it’s not the main thing. The main thing is that I want to help these students learn, and using AI is directly harmful to learning. Common sense says that prompting ChatGPT doesn’t have the same learning benefits as doing the reading and writing yourself, but if you need something more robust than common sense, here’s a sampling of the research literature.

…

There are two basic reasons for this and to make these clear I’m going to go way back to Augustine of Hippo’s ancient, uncompleted treatise De Musica.

https://wenatcheethehatchet.blogspot.com/2021/02/excerpts-from-augustines-de-musica-his.html 

I’ve quoted from Augustine extensively in the past because I have been annoyed at how he has been misrepresented (or misconstrued) by some pop-level depictions of him.  I have also taken some issue with philosophers like the late Roger Scruton concluding that since the neo-Platonic cosmology Augustine subscribed to has been discredited there is nothing to be gained from seriously dwelling on his ideas about music. 

Well, Scruton was wrong.  I’ll demonstrate Scruton was wrong by quoting from Augustine and showing that he articulated two basic points about the nature of music which are human capacities that should not just be farmed out to AI. Both of them are relatively obvious.  The first one is art is a product of human reason:

https://archive.org/details/augustine-on-music-de-musica 

Page 178

M. But don’t you think art is a sort of reason, and those who use art use reason? Or do you think otherwise?

 

D. It seems so.

 

M. Therefore, whoever cannot use reason does not use art. [emphasis added]

 

Page 179

M. … But if all imitation is art, and all art reason, all imitation is reason. But an irrational animal does not use reason; therefore, it does not possess an art. But it is capable of imitation; therefore art is not imitation.

 

Page 185

…

M. Then tell me, which is to be considered dearer, what is contained in our intellect or what is accidentally attributed to us by the judgment of an ignorant people?

 

D. No one doubts the first is far above all others, even those things which are not to be thought ours.

 

M. And so you don’t deny, do you, all science is contained in the intellect? [emphasis added]

 

D. Who does?

 

M. And, therefore, music is in the intellect.

 

D. That seems to follow from its definition. 

Whoever cannot use reason does not use art.  The first reason to avoid using AI to do your musical analytic thinking for you is, if you use it, you are abrogating the first and most important thing about music as an art, or any art, according to Augustine’s thoughts about what makes music an art to begin with. 

Is that an appeal to an authority?  Sure, but since hardly anyone has done much to improve upon Augustine’s observation in roughly sixteen centuries the appeal to authority isn’t invalid or unsound in this case. 

Now, we get to the second thing, which will take a bit longer to unpack: 

https://wenatcheethehatchet.blogspot.com/2021/02/augustines-de-musica-book-vi-necessity.html

 

Book VI, Chapter 3

Chapter 3

 

M. Notice, then, this third kind, being in the practice and operation of the person pronouncing, and see whether these numbers can be without those in the memory. For silent within ourselves we can also by thinking go through certain numbers in the amount of time they would be gone through by the voice. It is evident these are in a certain operation of the mind which, since it produces no sound and visits no passion on the ear, shows this kind of number can be without the other two, namely, the one in the sound, the other in the hearer when he hears. But we ask if it would be without memory’s accompanying it. Yet, if the soul produces the numbers we find in the beat of the veins, the question is solved.

 

Pages 329-330

M. Consider, too, the fourth class, that is, the class of those numbers in the memory. [emphasis added] For, if we draw them out by recollection, and, when we are carried away to other thoughts, we again leave them as if hidden in their own hiding places, I don’t think it is difficult to see they can be without the others.

 

D. I don’t doubt they can be without the others. But just the same, unless they were heard or thought, they could not be sent on to the memory.  And so, although they remain at the death of those that are heard or thought, yet they are imprinted by them.

 

M. I don’t contradict you, and I should like now to ask which of these four kinds you judge the principal one. Except, I believe, while we were discussing these things, a fifth kind appeared from somewhere, a kind in the natural judgment of perceiving when we are delighted by the equality of numbers or offended at a flaw in them. For I am mindful of your opinion our sense could have in no way done this without certain numbers latent in it. Or do you, perhaps, think a great power like this belongs to some one of those four?

 

D. On the contrary. I think this kind is to be distinguished from all of them. For it is one thing to sound and this is attributed to a body; another to hear, and in the body the soul is passive to this from sounds; another to produce numbers either more slow or less so; another to remember them; and another by accepting or rejecting, to give sentence on them all as if by some natural right.

 

M. Come, now, tell me which of these five is the most excellent.

 

D. The fifth, I think.

 

M. You are right, for, unless it excelled, it could not bring judgment on them. But again, I want to know of the other four which you judge the greatest.

 

D. The kind in the memory, certainly.  …

 

You may have all the reasoning powers in the world, yet if you can’t remember what you just heard you can’t use your reason to hear it as music.  Augustine laid out two simple aspects of human existence through which we hear music as music, we use our reason and we rely upon our memory. 

 

There have been, of course, all kinds of debates about those two points and how they apply to music.  My point is that if you rely on artificial intelligence to get your work done for you the you are abrogating the use of your power of reason, first of all, and second of all, you are abrogating your capacity to rely on memory as a musical resource. 

 

A few years ago I came across a situation where someone fed a post I wrote into an AI. I can’t remember which one it was but I remember it was a post I published about Matiegka’s use of the V⁷/V to V to IV progression in his guitar sonatas.  What little I can remember is the AI was catastrophically wrong.  It was as though what the LLM was good at doing was paraphrasing written words.  The trouble was that the written words were explaining analytical and theoretical work that was put into a score.  An AI that can’t read a score that has in-score observations and which exists as a JPEG file is only going to be able to paraphrase words, not observations in a score.

 

AI “could” hear how Robert Johnson’s “Kind-Hearted Woman” and Matiegka’s Grand Sonata II, 1 both feature V-IV regressions, but AI has not necessarily been programmed to listen across urbane 20^th century American blues and early 19^th century Bohemian guitar sonatas to hear a gesture that is shared across the two eras and styles of guitar music. 

 

The AI summary I read of my old post seemed to be a disaster.  It was as if the result was just a mishmash of surmising that I said Matiegka was innovative and inventive.  I wrote that he was considered conservative and conventional. What I found curious was how this conventional composer used V-IV regressions repeatedly in a  way that reminded me of a harmonic pattern that shows up all over blues songs. 

 

I have tried making this point before but I would say there’s a third reason to refrain from using AI in analytical work besides the fact that if you use AI you’re abdicating your own powers of reason and not cultivating your own musical memory; the AI is likely to give you a words-only explanation of a the textual reception history of a work which may only replicate scholastic super-myths that need to be debunked, or you may find that AI has nothing to say about a work because it’s too obscure for anyone to have plugged in a coherent opinion about it.  Let’s take Matiegka, there’s hardly any English-language writing about his work and who would feed the Czech or German scholarship into an AI model for American music students to benefit from? AI could end just feeding you a pre-fab myth about something musical that misleads you rather than informs you about what happens (or happened) in the music.

 

When I discovered Matiegka’s V-IV usage across his guitar sonatas it told me that it’s a myth perpetuated on both the classical and blues sides of music theory and history that blues broke all the rules of classical music (on the blues side) and that in classical music you “never” use a V-IV chord progression.  It’s the kind of observation that can’t (yet) be fed into an LLM for an AI pattern and it’s the kind of observation that a human can make using reason and memory.

 

It’s not that I don’t think it’s possible for AI to think across genres, what I am saying is that if you’re someone like me who has spent thirty years looking for potential points of convergence between ragtime and blues with the syntactic procedures of 18^th century music (i.e. a fusion of Haydn and Hooker, a synthesis of Joplin and Bach), AI has been programmed to spit out conventional wisdoms from within genres rather than get used (or asked?) to consider points of convergence and inter-stylistic or inter-formal fusion.

 

Your ability to reason and your musical memories are resources you should be seeking to cultivate as much as you can and not farming out to AI.  Why rack up student debt to get a music credential, for instance, if you at some point farm out your capacities of musical reasoning and musical memory to an AI just to knock out an assignment? 

 

I agree with Ethan Hein that reliance on AI inhibits the learning process.  Augustine’s De Musica is a slog to read but he articulated some very simple points about what it is about music that is human, and what it is humans have that enable them to hear something as music and remember it as music later.  AI usage is farming out to an LLM the most essential aspects of musical learning.  Maybe once you’ve crossed some threshold of musical competency AI usage won’t be a problem, but if you’re a beginner in music the likelihood that AI usage will mislead you rather than help you still seems high.

 

Some thoughts for the weekend, such as they are.