What a lovely and intelligent essay, to which I will append a deeply stupid example.
Many statistical inference algorithms are based on drawing repeated samples from a distribution. You only know what the distribution is like by looking at the samples. A bad sampling algorithm will get stuck basically sampling the point over and over again. It will look like it drew 100 samples, but actually you only got, like, 2 pieces of useful information.
A good sampling algorithm will sample independently. The next sample has nothing to do with with the last sample. If it looks like it drew 100 samples, you got 100 pieces of information.
A GREAT sampling algorithm will draw samples that are *anti-correlated*. It will draw 1 sample, then it will fly off into the distance and draw the next sample from as far away as it can go. If it looks like it drew 100 samples, you actually got like 200 pieces of information.
Shakespeare, at his best, writes like a GREAT sampling algorithm. Every idea is attacked from 20 different directions. The full space of interpretations, associations, implications is smashed into a few dozen lines. Sometimes it looks like it is losing coherence, losing its grip on the original problem; but no, actually, the problem is just much bigger and much weirder than it first seemed.
Great comment, Seth. It seems to tie into something I have been writing about- specifically that to inhabit language is to inhabit a space of questions.
The artistic director of a repertory theater I thrived in as a young actor once said, during a Saturday morning acting class he was coaching, in reference to the slog that is learning through rehearsal who a character is and why they say and do what they do, "Until the answers come (in any form of creativity through rehearsals/rewriting/sketching shapes before painting, et al) you have to embrace the notion of being comfortable living within the question."
That was sixty years ago. I've never forgotten that simple wisdom of' living within the question'. And have applied it to so many challenges in my life, and the lives of those I have loved.
The AI creations I like and that I think could approach art are those that “go off”—that hallucinate, make things up and then freak out when you call them on it, display that creepy obsequiousness or arrogance, or get pissy or bitchy, or are deeply uncanny—all qualities developers hasten to smother. But they’re revealing, funny, scary—and surprising.
One example i can think of off the top of my head is Beethoven's Ode to Joy (aka the hymn Joyful, Joyful We Adore Thee). For the first two lines it is a very stepwise series of notes. At the end of the third line, there is a large leap downward, which you would not expect after so many small changes in note value.
And I've also seen it in dynamics. The music gets softer, when you'd expect it to be louder, and vice versa.
Vollmann writes in “Europe Central”: “Every word, right down to its gaping letters o and grinning letters e, must offer resonance with sentences before and beyond—not predictability, mind you, for that would be tedious, but after each comma the hindsighted reader needs to be in the position of saying: Why didn’t I see that coming?”
I really enjoyed this essay, Nabeel—thank you for sharing your very eloquent thoughts.
I’d consider a fourth point, which I believe follows from the three you've already mentioned: part of what makes "great" art meaningful to us is our sense that it was hard for the artist to bring all its great elements together. The struggle of making it gives it a considerable part of its value. AI art, since it's designed to be created with "ease", lacks this necessary element of struggle, and feels inferior even if the output may appear compelling.
This is important in so much of human activity. We watch humans play chess as hard as they can even though a robot could beat them easily. We watch humans jump as far as they can as hard as they can even though a cheetah could beat them easily.
I think information has something to do with this as well. We are interested to know what the limits are of human capability; to understand what is possible with bodies like ours, with minds like ours. Or maybe we want to understand what is so interesting to our fellow humans about chess and/our jumping.
This is the best post I've ever read on AI writing. You put words to things I have only "felt" in the past couple of years. I've spent countless hours reading human-written content online since I was 10, so the transition to a world where most content is written by AI (or with its help) has been very clear to me. I couldn't have described it as eloquently as you did here, but outside of the obvious tell-tale signs (it's not X, it's Y, em-dashes...), there's a loss in personality that is immediately recognizable.
In my eyes, AI writing tries to cram a punchline or a thoughtful idea in short sentences that follow each other to maniacally offer "brand new ideas". Even after cleaning up the "it's not X, it's Y", what remains still tries really hard to punch you in the face with how much it knows, how many hidden links it uncovers. It loves to draw lines through complex systems, acknowledging their complexity but still retaining a verifiable outcome. I think that's a hallmark of what laypeople (myself included!) used to call "smart writing" a few years ago.
Writers who could deconstruct complex topics into causes and effects, and walk us through the system-wide movements instead of focusing too much on any one variable, seemed to have cracked the code. Now that AI writes like that for everybody, including those who don't understand the complexity of what they're writing about, the style itself went from a sign of value to a sign of a commoditized product.
Nabeel, this is excellent. I am a humanities/literature teacher. I am excited to have my students work through some of your concepts here. Also, about ten years ago my wife and I hosted an exchange student from Pakistan. We found your resources really helpful in our conversations with her. It was fun to rediscover some of your work in this area!
1. Your assumption is that AI is capable of making great art in the future. Maybe so. But I am not convinced society will care for AI-generated art. (I don't as much though I try not to intentionally discriminate.) Watch this Brandon Sanderson keynote on AI art, where he makes the argument that art is intrinsically human-oriented and we will not value AI art. (https://youtu.be/mb3uK-_QkOo)
Gemini summary of his arguments:
In this keynote, Brandon Sanderson argues that the true value of art lies not in the final product, but in the transformative process of creation (10:44 - 15:42). He posits that while AI can generate polished results, it cannot experience growth, learning, or genuine care, thereby stripping creators of the essential journey that shapes them into artists (15:42 - 16:55). Ultimately, Sanderson urges society to reclaim its agency by choosing to value human effort, asserting that because we define what art is, we must reject the commodification of creativity in favor of the intrinsic human experience (17:15 - 18:26). Sanderson ultimately concludes that we make art because we are driven to do so by our nature. He declares, "You are the art," urging creators to recognize that the value of creative expression lies in the human experience itself.
2. Eloquence has been plummeting in both quantity and status. So even if we did have great written AI art, I am not sure it would matter all that much. It would be the proverbial tree that falls in a forest and no one is around to hear it. https://www.nytimes.com/2026/04/02/opinion/eloquence-ai-arguments.html
3. This is mostly about written AI art, and I liked it. I'd like an essay centered on why AI video and AI movies are terrible.
My favorite line in poetry is Mary Oliver’s “Tell me about despair, yours” which is such a surprising construction (you’d normally say “tell me about your despair”). It packs so much punch bc it entreats the reader to turn the abstract into the personal (no, not just any despair, *your despair*).
I enjoyed the close readings of Shakespeare in this essay, which makes me want to revisit the Sonnets. Your list of "ear" references in HAMLET reminds me of another, vivid line from the play that I've always remembered: In Act III, after Hamlet chastises Gertrude for her remarriage, she says, "These words like daggers enter in mine ears."
Maybe uniqueness or idiosyncracy is better than surprise? Also your definition is pretty similar to C.S. Lewis's, that good art is that which gets better the closer you interrogate it. Bad art is the opposite - the closer you look, the more you notice the imperfections.
Ah, the western Canon, running unabashedly over the artifacts of history and culture that emanate from other places. It's so reassuring that William Shakespeare continues to be a primary tool for guarding the status quo.
An interesting article about inventiveness, nonetheless it does not show what AI can't or won't do, but more of what (mostly white male) humans can. It's not us against machines, it never was. Our children, and inventions will surpass the hamster on a wheel in a cage, even if the momentum continues long after it's run the course.
Thank you Nabeel for your essay. I'm an artist and a poet by my God-given DNA and my earned BFA but spent my entire adult life in a commercial career for more than 28 years. It was not until a recent divorce, career loss, and a neurodiverse aspie girl diagnosis that I picked up a paint brush again to work through the grief and anguish that became my every day. I always knew from a young age there was something special and layered about the art I create but did not have the language or knowledge to engage on the world stage. The people in my life encouraged me to do things so I would remain unnoticed. I am finally beginning to acknowledge my creative gifts and appreciate the time you took to post your humankindedness thoughts and peace in this world of AIROBOTICS.
Bravo, so much food for thought here. It feels like there's a relational piece here too, where we need to see the art object as mediating a relationship between creator and recipient and beautiful in how successfully it transcends the limits of ordinary communication ("if I could have said it plainly I wouldn't have written it as a poem"); the pleasure I get in playing a piece by Chopin is in imagining a sickly and depressive Pole in France who died young and who I'm connecting with, subliminally, transcending death and across two centuries, in moving my body as he would have moved his body, and making the phrases breathe as he would have made them breathe, hearing the intent behind what he wrote.. Even if a computer generated piece contains similar elements of surprise-under-constraints, has similarly beautiful phrasing or development, it's not solving that relational problem, it's maxxing a different reward function
Hi Nabeel, I've been approaching taste more top-down through a framework, and seeing your real examples filled in a lot for me. You've addressed most of the 10 dimensions I use here; I found particularly insightful was your 'surprise' section – I just labelled that as 'creativity' (subversion of rules/norms in medium+context), but your detail is really much sharper than mine.
Also reached a similar conclusion to you re: AI – until an AI can experience life as we do (but also will likely be a minor worry at that point). You may also be interested in Italo Calvino's 1991 definitions for a 'Classic book' :)
What a lovely and intelligent essay, to which I will append a deeply stupid example.
Many statistical inference algorithms are based on drawing repeated samples from a distribution. You only know what the distribution is like by looking at the samples. A bad sampling algorithm will get stuck basically sampling the point over and over again. It will look like it drew 100 samples, but actually you only got, like, 2 pieces of useful information.
A good sampling algorithm will sample independently. The next sample has nothing to do with with the last sample. If it looks like it drew 100 samples, you got 100 pieces of information.
A GREAT sampling algorithm will draw samples that are *anti-correlated*. It will draw 1 sample, then it will fly off into the distance and draw the next sample from as far away as it can go. If it looks like it drew 100 samples, you actually got like 200 pieces of information.
Shakespeare, at his best, writes like a GREAT sampling algorithm. Every idea is attacked from 20 different directions. The full space of interpretations, associations, implications is smashed into a few dozen lines. Sometimes it looks like it is losing coherence, losing its grip on the original problem; but no, actually, the problem is just much bigger and much weirder than it first seemed.
Your points reminded me of this Teju Cole essay arguing that good photography needs to be surprising and contrasting it with the kind of superficially and formally “beautiful” photographs that are nevertheless “predictable”: https://www.nytimes.com/2016/04/03/magazine/a-too-perfect-picture.html?unlocked_article_code=1.hVA.Rej6.rS0umYzd9hM0&smid=nytcore-ios-share (The predictable photos probably resemble what a generative AI model would give you.)
Great comment, Seth. It seems to tie into something I have been writing about- specifically that to inhabit language is to inhabit a space of questions.
The artistic director of a repertory theater I thrived in as a young actor once said, during a Saturday morning acting class he was coaching, in reference to the slog that is learning through rehearsal who a character is and why they say and do what they do, "Until the answers come (in any form of creativity through rehearsals/rewriting/sketching shapes before painting, et al) you have to embrace the notion of being comfortable living within the question."
That was sixty years ago. I've never forgotten that simple wisdom of' living within the question'. And have applied it to so many challenges in my life, and the lives of those I have loved.
Thank you, Jim. That’s a great memory!
The AI creations I like and that I think could approach art are those that “go off”—that hallucinate, make things up and then freak out when you call them on it, display that creepy obsequiousness or arrogance, or get pissy or bitchy, or are deeply uncanny—all qualities developers hasten to smother. But they’re revealing, funny, scary—and surprising.
have you seen it in music?
Yes, it exists in music. Have you studied music theory?
naur
One example i can think of off the top of my head is Beethoven's Ode to Joy (aka the hymn Joyful, Joyful We Adore Thee). For the first two lines it is a very stepwise series of notes. At the end of the third line, there is a large leap downward, which you would not expect after so many small changes in note value.
And I've also seen it in dynamics. The music gets softer, when you'd expect it to be louder, and vice versa.
Vollmann writes in “Europe Central”: “Every word, right down to its gaping letters o and grinning letters e, must offer resonance with sentences before and beyond—not predictability, mind you, for that would be tedious, but after each comma the hindsighted reader needs to be in the position of saying: Why didn’t I see that coming?”
I really enjoyed this essay, Nabeel—thank you for sharing your very eloquent thoughts.
I’d consider a fourth point, which I believe follows from the three you've already mentioned: part of what makes "great" art meaningful to us is our sense that it was hard for the artist to bring all its great elements together. The struggle of making it gives it a considerable part of its value. AI art, since it's designed to be created with "ease", lacks this necessary element of struggle, and feels inferior even if the output may appear compelling.
This is important in so much of human activity. We watch humans play chess as hard as they can even though a robot could beat them easily. We watch humans jump as far as they can as hard as they can even though a cheetah could beat them easily.
I think information has something to do with this as well. We are interested to know what the limits are of human capability; to understand what is possible with bodies like ours, with minds like ours. Or maybe we want to understand what is so interesting to our fellow humans about chess and/our jumping.
This is the best post I've ever read on AI writing. You put words to things I have only "felt" in the past couple of years. I've spent countless hours reading human-written content online since I was 10, so the transition to a world where most content is written by AI (or with its help) has been very clear to me. I couldn't have described it as eloquently as you did here, but outside of the obvious tell-tale signs (it's not X, it's Y, em-dashes...), there's a loss in personality that is immediately recognizable.
In my eyes, AI writing tries to cram a punchline or a thoughtful idea in short sentences that follow each other to maniacally offer "brand new ideas". Even after cleaning up the "it's not X, it's Y", what remains still tries really hard to punch you in the face with how much it knows, how many hidden links it uncovers. It loves to draw lines through complex systems, acknowledging their complexity but still retaining a verifiable outcome. I think that's a hallmark of what laypeople (myself included!) used to call "smart writing" a few years ago.
Writers who could deconstruct complex topics into causes and effects, and walk us through the system-wide movements instead of focusing too much on any one variable, seemed to have cracked the code. Now that AI writes like that for everybody, including those who don't understand the complexity of what they're writing about, the style itself went from a sign of value to a sign of a commoditized product.
Nabeel, this is excellent. I am a humanities/literature teacher. I am excited to have my students work through some of your concepts here. Also, about ten years ago my wife and I hosted an exchange student from Pakistan. We found your resources really helpful in our conversations with her. It was fun to rediscover some of your work in this area!
1. Your assumption is that AI is capable of making great art in the future. Maybe so. But I am not convinced society will care for AI-generated art. (I don't as much though I try not to intentionally discriminate.) Watch this Brandon Sanderson keynote on AI art, where he makes the argument that art is intrinsically human-oriented and we will not value AI art. (https://youtu.be/mb3uK-_QkOo)
Gemini summary of his arguments:
In this keynote, Brandon Sanderson argues that the true value of art lies not in the final product, but in the transformative process of creation (10:44 - 15:42). He posits that while AI can generate polished results, it cannot experience growth, learning, or genuine care, thereby stripping creators of the essential journey that shapes them into artists (15:42 - 16:55). Ultimately, Sanderson urges society to reclaim its agency by choosing to value human effort, asserting that because we define what art is, we must reject the commodification of creativity in favor of the intrinsic human experience (17:15 - 18:26). Sanderson ultimately concludes that we make art because we are driven to do so by our nature. He declares, "You are the art," urging creators to recognize that the value of creative expression lies in the human experience itself.
2. Eloquence has been plummeting in both quantity and status. So even if we did have great written AI art, I am not sure it would matter all that much. It would be the proverbial tree that falls in a forest and no one is around to hear it. https://www.nytimes.com/2026/04/02/opinion/eloquence-ai-arguments.html
3. This is mostly about written AI art, and I liked it. I'd like an essay centered on why AI video and AI movies are terrible.
My favorite line in poetry is Mary Oliver’s “Tell me about despair, yours” which is such a surprising construction (you’d normally say “tell me about your despair”). It packs so much punch bc it entreats the reader to turn the abstract into the personal (no, not just any despair, *your despair*).
I enjoyed the close readings of Shakespeare in this essay, which makes me want to revisit the Sonnets. Your list of "ear" references in HAMLET reminds me of another, vivid line from the play that I've always remembered: In Act III, after Hamlet chastises Gertrude for her remarriage, she says, "These words like daggers enter in mine ears."
This is great.
It seems that the link to the interactive essay is broken, though.
thanks fixed!
Maybe uniqueness or idiosyncracy is better than surprise? Also your definition is pretty similar to C.S. Lewis's, that good art is that which gets better the closer you interrogate it. Bad art is the opposite - the closer you look, the more you notice the imperfections.
Ah, the western Canon, running unabashedly over the artifacts of history and culture that emanate from other places. It's so reassuring that William Shakespeare continues to be a primary tool for guarding the status quo.
An interesting article about inventiveness, nonetheless it does not show what AI can't or won't do, but more of what (mostly white male) humans can. It's not us against machines, it never was. Our children, and inventions will surpass the hamster on a wheel in a cage, even if the momentum continues long after it's run the course.
Thank you Nabeel for your essay. I'm an artist and a poet by my God-given DNA and my earned BFA but spent my entire adult life in a commercial career for more than 28 years. It was not until a recent divorce, career loss, and a neurodiverse aspie girl diagnosis that I picked up a paint brush again to work through the grief and anguish that became my every day. I always knew from a young age there was something special and layered about the art I create but did not have the language or knowledge to engage on the world stage. The people in my life encouraged me to do things so I would remain unnoticed. I am finally beginning to acknowledge my creative gifts and appreciate the time you took to post your humankindedness thoughts and peace in this world of AIROBOTICS.
Bravo, so much food for thought here. It feels like there's a relational piece here too, where we need to see the art object as mediating a relationship between creator and recipient and beautiful in how successfully it transcends the limits of ordinary communication ("if I could have said it plainly I wouldn't have written it as a poem"); the pleasure I get in playing a piece by Chopin is in imagining a sickly and depressive Pole in France who died young and who I'm connecting with, subliminally, transcending death and across two centuries, in moving my body as he would have moved his body, and making the phrases breathe as he would have made them breathe, hearing the intent behind what he wrote.. Even if a computer generated piece contains similar elements of surprise-under-constraints, has similarly beautiful phrasing or development, it's not solving that relational problem, it's maxxing a different reward function
Hi Nabeel, I've been approaching taste more top-down through a framework, and seeing your real examples filled in a lot for me. You've addressed most of the 10 dimensions I use here; I found particularly insightful was your 'surprise' section – I just labelled that as 'creativity' (subversion of rules/norms in medium+context), but your detail is really much sharper than mine.
Also reached a similar conclusion to you re: AI – until an AI can experience life as we do (but also will likely be a minor worry at that point). You may also be interested in Italo Calvino's 1991 definitions for a 'Classic book' :)
Thank you.