My husband and I have notoriously opposite musical tastes, and you in fact helped us find overlap! I had a feeling he’d like this too and he does, in fact he loves it so much he has already listened to it multiple times 😄 he’s a guitarist and base player himself and typically into math metal, but the complexity of this really tickles his brain. So thanks for that as well!
Thanks for the music comments. Same for the homogenization of music by AI. All the pablum on Spotify and other channels generated by AI are derivative and unexciting, unlike the brilliant and original work you showcased here. There is a narrative in music, even acoustic music and the colored blob for someone like Yasmin Williams would have a huge balloon at the top compared to the AI ovals.
Love the blobs. I think the trophy for the homogenization of literature goes to Amazon kindle. With Grammarly running a close 2nd. But that is a point for Ai. Just sayin'
Fascinating study on AI vs. human narratives. The following (very recent) study makes a similar point, based on an analysis of human +1500 responses from New York Times and Boston Review forum debates - compared with 23,384 LLM-generated essays: https://arxiv.org/html/2606.01736v3.
At the overall level of argument, 65.3% of human main arguments were unique within a given debate, compared with only 3.4% of GenAI-generated main arguments. The arguments GenAI provided might not be bad, but far less diverse than what human thinkers offer. Furthermore, human arguments are typically more concrete and grounded in specific cases or interventions. In contrast, GenAI arguments tended to be more abstract, rely on generic frameworks, mention fewer specific cases - and hedge far more, using general statements (see image to the right below). Finally, at the structural level GenAI models also tended to collapse. GenAI essays were more likely to follow the same structure (claim, support, proposal) - while human written essays tended to vary more in their structure and ways to build an argument.
Who knows how this might play out over the following years - in terms of the risk of homogenization of human thinking and writing. I do think high schools (e.g.) have an excellent opportunity to insist on writing tasks in class, that ensure that people get to try to write - alone.
Ooh, I hadn't seen that paper, that's a good one. I think this is a general feature even outside of writing. This paper suggest homogenization, in a sense, of science: https://www.nature.com/articles/s41586-025-09922-y
also agree that this offers a unique opportunity for secondary schools - where there is quite a lot of control over how kids time is spent - to emphasize and protect writing & thinking without machine assistance
I regularly wrestle with these questions of originality and AI composition, but I operate from the point of view of somebody who was educated in a one-room country schoolhouse three-quarters of a century ago. (see my direct response to the article, above.) That means I have written for a long time, when there wasn't much question of relying on an outside force to write school essays. You might parrot other human thoughts, but you still had to write the words down yourself, no cut and paste.
I do appreciate that educators have to battle to prevent kids from skating through. But as a contrarian, my interest turns to those few students who are accused of plagiarism when they have provably done their own work. That's the dilemma: years ago, my seventh-grade daughter was accused of exactly that when she turned in a 60-page "original" novel. The teacher sternly told her she would only get a B because it was obvious that her mother, a NYT-bestselling novelist, had helped with the composition. We never heard about that injustice until years later, long after our daughter had decided not to become a writer.
Tim Requarth: you are filling a gap in the Substack community re AI and education. Good on 'ya, as the Aussies say. I value that because in my recent conversion (sometimes I feel like one of the early Christians), I need all the eductation I can get. I'm trying to fill another gap, introducing older folks who may be interested in AI but feel baffled by it. We are working on adjacent gaps and hopefully, we can patch a hole in the web of the new world.
The good thing is that both of us understand how vital education is going to be, if AI is going to move forward. There are plenty of people out there who seem bent on burning down data centers and discouraging all AI use. Not all of those people have clean motives. (See Data republican on Substack for further information on that).
For a contrary old cuss like me, Substack needs to include a new button: "Like, but..."
Tim Requarth's piece on originality is well-worth reading, BUT...as somebody who has scribbled for a living and for pleasure for more than 74 years, I feel that originality all by itself is a hollow standard for judging creative work. A truly "original" work is like Phillip Glass music, solipsistic and inbred (to me. That always goes almost without saying.) Original work is in danger of being incomprehensible to anyone except the author. I say that because I have written stuff that fell into that category.
Artists of all kinds are thieves. They exist within a culture that can embrace their work or not. There are new stories, but ultimately no new story lines, except for the occasional genius who usually is not acknowledged until centuries later, and then only by doctoral candidates seeking dissertation topics.
Granted, that means artists can become rote hacks, but frankly we don't need to blame AI for "slop." Humans produce plenty of that all by their lonesome selves. Animation of story lines with new characters and new stage dressing touches is a form of originality every bit as welcome, maybe more welcome, than "original" work that is accomplished by standing every convention on its head.
In fact, I would say in my doddering old age that there will soon be an LLM named "Topsy-Turvey" that reverses every accepted plot twist and character change in our present story-telling manual.
And I'll bet some critics and authors will applaud and adopt it.
So there! One of the real joys of old age is being able to say what you mean and hope that it sparks a brouhaha that we all can use as an excuse to jump in and discuss. Not flame war but discussion.
Now, let me go back and listen to the guitar recordings.
Thank you for sharing that gorgeous piece of music at the end!
He’s really a quite phenomenal guitarist…so glad you enjoyed
My husband and I have notoriously opposite musical tastes, and you in fact helped us find overlap! I had a feeling he’d like this too and he does, in fact he loves it so much he has already listened to it multiple times 😄 he’s a guitarist and base player himself and typically into math metal, but the complexity of this really tickles his brain. So thanks for that as well!
Thanks for the music comments. Same for the homogenization of music by AI. All the pablum on Spotify and other channels generated by AI are derivative and unexciting, unlike the brilliant and original work you showcased here. There is a narrative in music, even acoustic music and the colored blob for someone like Yasmin Williams would have a huge balloon at the top compared to the AI ovals.
I feel like this is why that wild Montreal band Angine de Poitrine went viral recently, they basically musically personify the top of the blob
Didn’t know them so I checked Spotify. Thought , “That’s interesting but they couldn’t possibly play that Live.” So I checked YouTube, OMG
https://youtu.be/0Ssi-9wS1so?is=fahatNn8YD8TzCgg
Love the blobs. I think the trophy for the homogenization of literature goes to Amazon kindle. With Grammarly running a close 2nd. But that is a point for Ai. Just sayin'
lol totally agree that other homogenizing forces precede the latest wave of AI. I loved this post: https://kpb12177.substack.com/p/your-ai-voice-has-a-google-problem
Great article!
Fascinating study on AI vs. human narratives. The following (very recent) study makes a similar point, based on an analysis of human +1500 responses from New York Times and Boston Review forum debates - compared with 23,384 LLM-generated essays: https://arxiv.org/html/2606.01736v3.
At the overall level of argument, 65.3% of human main arguments were unique within a given debate, compared with only 3.4% of GenAI-generated main arguments. The arguments GenAI provided might not be bad, but far less diverse than what human thinkers offer. Furthermore, human arguments are typically more concrete and grounded in specific cases or interventions. In contrast, GenAI arguments tended to be more abstract, rely on generic frameworks, mention fewer specific cases - and hedge far more, using general statements (see image to the right below). Finally, at the structural level GenAI models also tended to collapse. GenAI essays were more likely to follow the same structure (claim, support, proposal) - while human written essays tended to vary more in their structure and ways to build an argument.
Who knows how this might play out over the following years - in terms of the risk of homogenization of human thinking and writing. I do think high schools (e.g.) have an excellent opportunity to insist on writing tasks in class, that ensure that people get to try to write - alone.
Ooh, I hadn't seen that paper, that's a good one. I think this is a general feature even outside of writing. This paper suggest homogenization, in a sense, of science: https://www.nature.com/articles/s41586-025-09922-y
also agree that this offers a unique opportunity for secondary schools - where there is quite a lot of control over how kids time is spent - to emphasize and protect writing & thinking without machine assistance
I regularly wrestle with these questions of originality and AI composition, but I operate from the point of view of somebody who was educated in a one-room country schoolhouse three-quarters of a century ago. (see my direct response to the article, above.) That means I have written for a long time, when there wasn't much question of relying on an outside force to write school essays. You might parrot other human thoughts, but you still had to write the words down yourself, no cut and paste.
I do appreciate that educators have to battle to prevent kids from skating through. But as a contrarian, my interest turns to those few students who are accused of plagiarism when they have provably done their own work. That's the dilemma: years ago, my seventh-grade daughter was accused of exactly that when she turned in a 60-page "original" novel. The teacher sternly told her she would only get a B because it was obvious that her mother, a NYT-bestselling novelist, had helped with the composition. We never heard about that injustice until years later, long after our daughter had decided not to become a writer.
I agree that questions of originality are often more complex than people like to admit…appreciate you reading and your thoughts!!
Tim Requarth: you are filling a gap in the Substack community re AI and education. Good on 'ya, as the Aussies say. I value that because in my recent conversion (sometimes I feel like one of the early Christians), I need all the eductation I can get. I'm trying to fill another gap, introducing older folks who may be interested in AI but feel baffled by it. We are working on adjacent gaps and hopefully, we can patch a hole in the web of the new world.
The good thing is that both of us understand how vital education is going to be, if AI is going to move forward. There are plenty of people out there who seem bent on burning down data centers and discouraging all AI use. Not all of those people have clean motives. (See Data republican on Substack for further information on that).
For a contrary old cuss like me, Substack needs to include a new button: "Like, but..."
Tim Requarth's piece on originality is well-worth reading, BUT...as somebody who has scribbled for a living and for pleasure for more than 74 years, I feel that originality all by itself is a hollow standard for judging creative work. A truly "original" work is like Phillip Glass music, solipsistic and inbred (to me. That always goes almost without saying.) Original work is in danger of being incomprehensible to anyone except the author. I say that because I have written stuff that fell into that category.
Artists of all kinds are thieves. They exist within a culture that can embrace their work or not. There are new stories, but ultimately no new story lines, except for the occasional genius who usually is not acknowledged until centuries later, and then only by doctoral candidates seeking dissertation topics.
Granted, that means artists can become rote hacks, but frankly we don't need to blame AI for "slop." Humans produce plenty of that all by their lonesome selves. Animation of story lines with new characters and new stage dressing touches is a form of originality every bit as welcome, maybe more welcome, than "original" work that is accomplished by standing every convention on its head.
In fact, I would say in my doddering old age that there will soon be an LLM named "Topsy-Turvey" that reverses every accepted plot twist and character change in our present story-telling manual.
And I'll bet some critics and authors will applaud and adopt it.
So there! One of the real joys of old age is being able to say what you mean and hope that it sparks a brouhaha that we all can use as an excuse to jump in and discuss. Not flame war but discussion.
Now, let me go back and listen to the guitar recordings.