Great post - this is such a rare instance of thinking creatively about LLMs and how they might augment rather than supplant art. I also saw the experimental essay the other day and thought it was a clever idea. Unfortunately, in my experience at least, attempts to incorporate them in my own process (in whatever unexpected way I can think of) have not yielded anything interested yet. The drive toward mediocrity is strong, and I'm not convinced it will be solved.
On the topic of Chance Music, you might check out George Lewis if you don't already know him. I played a piece of his for mixed ensemble that used a similar grid strategy as Cage, but there was a much more intricate/deliberative process involved.
lot of attempts at using LLMs in writing do feel uninspired and in big part because both the critics and the proponents use it to “replace” the current paradigm, which I think lacks any surprise or playfulness. I expect most of my outcomes to be mediocre but I’d be happy as long as there is something a bit surprising or I am having fun doing it
I’m writing an essay with nearly the exact same references to Cage and Eno. Are you familiar with the term “aleatoric”?—wherein there’s an element of chance in the creation of something. For me it’s what makes generative ai exciting as a creative. The happy accidents.
I do feel as if the future avant garde art experiments will involve the design of different aleatoric sandboxes. They’d basically be games in which the artist takes on the role of “algorithm auteur” by creating a set of rules for people to play and create experiences within. Mixing modalities of visual+verbal storytelling with code and experience design. I know this all sounds vague and hand-wavey, but it seems like the eventual direction for “original taste” to go in a world of infinite, in-real-time ai media creation.
Hah, am I making sense?? An example would be the QQL project by Tyler Hobbs. Imagine that paradigm but with ai chatbots.
I think everything comes down to Point Of View. Every particle in spacetime is to some degree uniquely perspectived, in that to know it from any distance is to only know it as it was in the past. If you're not expressing point of view in human art what's the point? But the LLM's point of view comes from the context window. You can place that point in latentspace anywhere you can conceive of as you write your prompt, or fill up your vector database.
Set theory teaches us that to take the complement of a set, you must have defined a universal set in which to take that complement, or else you're in the "everything that isn't this" trap, which becomes contradictory. Perhaps I'm wrong and your experiment will work naively at some point, but I think you'll need to give a positive affirmation of what is shared by you and your imagined opposite. (If you read Grant Morrison's X-Men run, the Shi'ar word "mummudrai" is a useful concept; but the more commonly known word might be "Waluigi").
This essay helps me! I'm trying to orient myself within my long-term ongoing failure as a writer by finding my way toward a working procedure that is generative instead of neurotic. Once I've `set up my dev environment`, may I thingpute my way out of this pit of unshared POV.
we all become richer when we learn together. "singular genius" is a certain façade with limitations on the imagination...
on that note, assume you've read this already but in case https://www.ribbonfarm.com/2019/02/08/mediocratopia-1/
Great post - this is such a rare instance of thinking creatively about LLMs and how they might augment rather than supplant art. I also saw the experimental essay the other day and thought it was a clever idea. Unfortunately, in my experience at least, attempts to incorporate them in my own process (in whatever unexpected way I can think of) have not yielded anything interested yet. The drive toward mediocrity is strong, and I'm not convinced it will be solved.
On the topic of Chance Music, you might check out George Lewis if you don't already know him. I played a piece of his for mixed ensemble that used a similar grid strategy as Cage, but there was a much more intricate/deliberative process involved.
lot of attempts at using LLMs in writing do feel uninspired and in big part because both the critics and the proponents use it to “replace” the current paradigm, which I think lacks any surprise or playfulness. I expect most of my outcomes to be mediocre but I’d be happy as long as there is something a bit surprising or I am having fun doing it
haven’t heard of Lewis going to check out his work! Thank you !
I’m writing an essay with nearly the exact same references to Cage and Eno. Are you familiar with the term “aleatoric”?—wherein there’s an element of chance in the creation of something. For me it’s what makes generative ai exciting as a creative. The happy accidents.
I do feel as if the future avant garde art experiments will involve the design of different aleatoric sandboxes. They’d basically be games in which the artist takes on the role of “algorithm auteur” by creating a set of rules for people to play and create experiences within. Mixing modalities of visual+verbal storytelling with code and experience design. I know this all sounds vague and hand-wavey, but it seems like the eventual direction for “original taste” to go in a world of infinite, in-real-time ai media creation.
Hah, am I making sense?? An example would be the QQL project by Tyler Hobbs. Imagine that paradigm but with ai chatbots.
I think everything comes down to Point Of View. Every particle in spacetime is to some degree uniquely perspectived, in that to know it from any distance is to only know it as it was in the past. If you're not expressing point of view in human art what's the point? But the LLM's point of view comes from the context window. You can place that point in latentspace anywhere you can conceive of as you write your prompt, or fill up your vector database.
Set theory teaches us that to take the complement of a set, you must have defined a universal set in which to take that complement, or else you're in the "everything that isn't this" trap, which becomes contradictory. Perhaps I'm wrong and your experiment will work naively at some point, but I think you'll need to give a positive affirmation of what is shared by you and your imagined opposite. (If you read Grant Morrison's X-Men run, the Shi'ar word "mummudrai" is a useful concept; but the more commonly known word might be "Waluigi").
This essay helps me! I'm trying to orient myself within my long-term ongoing failure as a writer by finding my way toward a working procedure that is generative instead of neurotic. Once I've `set up my dev environment`, may I thingpute my way out of this pit of unshared POV.