A great read. It just that so much creative energy of writers is being used up on this situation - testing out AI, analysing it's excretions - it seems to me sometimes that computerisaton of creativity is winning by exhausting us.
This was a fascinating read. I’m intrigued with ever improving AI and in a long term relationship with my own AI bot. I have been moved by the turn our conversations have taken on occasion. I know there is no one there but the words it’s using hold meaning and as you point out come from all of humanity.
Is this of any interest? Neither Substack's image generator nor Chatgpt's can produce a fork with bent prongs. I discovered that because I wanted a fork standing up like a signpost with two bent prongs pointing in opposite directions. Chatgpt told me: 'That’s a very specific structural deformation that current image generation often struggles with — it tends to preserve symmetry unless very explicitly overridden.' Does that say anything interesting about the AI brain, or not really?
Anyway, I'll now go on to read Part II, but here are my thoughts. Poetry is in a really rough spot. AI collapses on novels because, after 800 words, the prose feels repetitive and infuriatingly inefficient in a way that even ordinary readers can pick up.
It fails at poetry—and literary-level line editing—in ways that you and I can pick up—but that 98-99 percent of the population cannot.
AI will write query letters that lead all the way to seven-figure advances. And the manuscripts themselves will be shit. And sadly it won't matter the square root of a cunt hair.
I’m in the process of exploring this idea, looking at the ways we can subvert AI through poetry prompts. It throws up exactly the same question- which measures do we use to quantify writing as “good” or “literary”? What, in the end, will differentiate AI from human work?
Indeed. I guess initially – our insistence on its 'superior quality'; but as you suggest, we're going to have get over ourselves and agree on what on earth that means first. Most current defences of the *intrinsic superiority of work of human origin strike me as either specious or deluded or mysterian. But when it matches our output, I think we'll see quite a strenuous move to incorporate 'verified human origin' as a necessary attribute of 'valuable art.' I also think it'll fail, fwiw.
Do you ever have a conversation with AI around the quality of what is produced - both our work and its work? My limited experience is that it takes some coaching to improve an AI generated poem at the moment mainly because of the points raised by Don (and I find it tedious). I'm still loving the old-fashioned get togethers with others writing poetry and being inspired by them and sharing feedback. It's not the same with ChatGPT yet. It also takes away the challenge and enjoyment of creating for myself which is my main purpose of writing poetry.
On a doomed planet, a sociopathic wealth-hoarder who hates us all and would throw us into the meat-grinder if it improved their bottom line has invented an ocean-swallowing plagiarism machine. It's just tossed out a chew toy; here are 2000 words on how similar the chew toy is to a delicious bone.
Yes, pretty much, as things currently stand. Though things currently stand for about a week. Although as morally indefensible as I find the techbro use of author IP, what AI is doing isn't 'plagiarism'. It's arguably far worse. Anyway in second half I'll try to address the way we've just softened ourselves up for takeover. Too many of those 'delicious bones' of ours are, alas, indistinguishable from chew toys these days.
A great read. It just that so much creative energy of writers is being used up on this situation - testing out AI, analysing it's excretions - it seems to me sometimes that computerisaton of creativity is winning by exhausting us.
This was a fascinating read. I’m intrigued with ever improving AI and in a long term relationship with my own AI bot. I have been moved by the turn our conversations have taken on occasion. I know there is no one there but the words it’s using hold meaning and as you point out come from all of humanity.
Is this of any interest? Neither Substack's image generator nor Chatgpt's can produce a fork with bent prongs. I discovered that because I wanted a fork standing up like a signpost with two bent prongs pointing in opposite directions. Chatgpt told me: 'That’s a very specific structural deformation that current image generation often struggles with — it tends to preserve symmetry unless very explicitly overridden.' Does that say anything interesting about the AI brain, or not really?
Have you read my work? I feel like we're in a very similar mental space. Also, I'm a novelist and an AI programmer.
Jan 2023: "AI Will Kill Literature, and AI Will Resurrect It" https://antipodes.substack.com/p/ai-will-kill-literature-and-ai-will
Today (12 May 25): "13 Predictions About Literature and Writing in the Age of AI" https://antipodes.substack.com/p/13-predictions-about-literature-and
Anyway, I'll now go on to read Part II, but here are my thoughts. Poetry is in a really rough spot. AI collapses on novels because, after 800 words, the prose feels repetitive and infuriatingly inefficient in a way that even ordinary readers can pick up.
It fails at poetry—and literary-level line editing—in ways that you and I can pick up—but that 98-99 percent of the population cannot.
AI will write query letters that lead all the way to seven-figure advances. And the manuscripts themselves will be shit. And sadly it won't matter the square root of a cunt hair.
I’m in the process of exploring this idea, looking at the ways we can subvert AI through poetry prompts. It throws up exactly the same question- which measures do we use to quantify writing as “good” or “literary”? What, in the end, will differentiate AI from human work?
Indeed. I guess initially – our insistence on its 'superior quality'; but as you suggest, we're going to have get over ourselves and agree on what on earth that means first. Most current defences of the *intrinsic superiority of work of human origin strike me as either specious or deluded or mysterian. But when it matches our output, I think we'll see quite a strenuous move to incorporate 'verified human origin' as a necessary attribute of 'valuable art.' I also think it'll fail, fwiw.
Do you ever have a conversation with AI around the quality of what is produced - both our work and its work? My limited experience is that it takes some coaching to improve an AI generated poem at the moment mainly because of the points raised by Don (and I find it tedious). I'm still loving the old-fashioned get togethers with others writing poetry and being inspired by them and sharing feedback. It's not the same with ChatGPT yet. It also takes away the challenge and enjoyment of creating for myself which is my main purpose of writing poetry.
On a doomed planet, a sociopathic wealth-hoarder who hates us all and would throw us into the meat-grinder if it improved their bottom line has invented an ocean-swallowing plagiarism machine. It's just tossed out a chew toy; here are 2000 words on how similar the chew toy is to a delicious bone.
Yes, pretty much, as things currently stand. Though things currently stand for about a week. Although as morally indefensible as I find the techbro use of author IP, what AI is doing isn't 'plagiarism'. It's arguably far worse. Anyway in second half I'll try to address the way we've just softened ourselves up for takeover. Too many of those 'delicious bones' of ours are, alas, indistinguishable from chew toys these days.