So spot-on, Nick. And so beautifully (I might even say, theatrically) framed.
(I’m also having a Doh moment. That creative industries seminar of mine you visited in 2019 looked briefly at the book digitisation landscape. Would’ve been great to have added your view on these matters. Alas, I left that teaching gig that summer, and post-pandemic swim in other waters.)
Reading your Fragments keeps my mind attuned. Thank you.
"It’s an abstract of Rembrandt, as if he were taken out of his own life and made to work across it." There is absolutely, 100% an entire novel in that sentence.
With respect to your point about the language that's used to describe generative computing, I am seeing it leading people astray time and again. The term 'artificial intelligence' is so far from the truth that it makes me spit feathers. It leads people to anthropomorphise, and you see that in discussions of whether "AI is lying to you" (it's not, it can't lie, it's just wrong), or "bullshitting". Even the preferred term of "hallucinating" is implying a level of agency that just isn't there.
I wonder if we need to collectively discuss more appropriate terminology. I prefer to call it 'generative computing' where possible, but always have to explain what that is. Same with using the initialism LLM. Using 'AI' often comes with the subtext that the software under discussion is, or is about to become, sentient, and that then brings with it a set of assumptions about intention, rights and personality, and the creation of parasocial relationships, that are deeply troubling.
Blessed be those taking an intelligent approach framed in an international context (or any context that doesn't bipolarise bots between skynet and BS hype-scent). It would be interesting to extend your analysis into an East vs West treatment.
My two other main concerns beyond Americani[s|z]ation of IP are:
1) Filtering: Yes, chatbots will turn most writing into 'Instagram for words'. Worse, these bots will even be used to provide the initial prompts for creating awkward, statistically relevant, yet emotionally empty text. But, worst of all, these prompts will be pre-filtered into 'socially corrected' memes by those who see their job as protecting free speech +/or social harmony whilst simultaneously denying +/or inflicting both.
2) Revisionism:
Rewriting text to ensure eternal cultural validity will become a culture-washing role, primarily for institutionally owned publishers, but also socially acceding grafted writers, politicians will hardly hinder the erasure of culpable error. Not only feeding the unleashed AI beasts, we'll also help prune and burn their uncomfortable parts. Sifting through the digital ash from bot-driven revision, historians will never divine the truth from the Corpus of history. We'll never be allowed to know it. We'll all unwittingly become willing Winstons.
Maybe information wants to be free. Work always has a cost -- that's true in the marketplace and it's true in physics. My view: a bit of software that incorporates my work in its own products has to pay for the right.
Many thanks for being at the forefront of intellectual property in the neural network age. While I don't believe for a second that the current generation of AIs—ChatGPT, Bing or Google's Bard—have anything resembling actual intelligence, I do see that they copy existing art.
I can't even say 'shamelessly' as these neural networks are devoid of emotions like 'shame'. Their proprietors may have a semblance of integrity hidden somewhere deep inside, but who am I fooling? The moment those neural networks generate big bucks, the plagiarised artists are on their own, without the massive amount of lawyers the likes of Google, Microsoft and Meta have by the bucketload.
It makes me extremely reluctant to post even small parts of a WiP (work in progress) online, and wonder if I should publish on Google Play, as this means your work is automatically put on Google Books, and is subsequently training material for their Google Bard neural network.
It's a great point that you can't predict the next Rembrandt painting by averaging all that came before. One might even argue that his paintings diverged increasingly from being 'average Rembrandt' over time - and became more essentially Rembrandt (which aligns with the German IP approach). My go-to counter to the hype is, sure, your machine model can generate a cubist-look image - but can it invent cubism?
As always Nick, it's both your posts and the comments on here that are so valuable. Re-read EM Foster's "The Duty of Society to the Artist" essay this morning, and 'generative computing' seems to be the 'centralising and disappearing' of the artist that Forster feared.
AI as a derivative work is fascinating and rings... true. The foundation and meaning behind the different copyright and IP laws you shared is fascinating, and I would love to learn more and about more of them, what are or were other traditions of that? Have you read this? https://www.noemamag.com/ai-is-life
Yes we need UBI or whatever the global version of nationalization is or ownership by all of the creators, which as this noama article argues could be everyone, ever.
So spot-on, Nick. And so beautifully (I might even say, theatrically) framed.
(I’m also having a Doh moment. That creative industries seminar of mine you visited in 2019 looked briefly at the book digitisation landscape. Would’ve been great to have added your view on these matters. Alas, I left that teaching gig that summer, and post-pandemic swim in other waters.)
Reading your Fragments keeps my mind attuned. Thank you.
"It’s an abstract of Rembrandt, as if he were taken out of his own life and made to work across it." There is absolutely, 100% an entire novel in that sentence.
Really good piece!
With respect to your point about the language that's used to describe generative computing, I am seeing it leading people astray time and again. The term 'artificial intelligence' is so far from the truth that it makes me spit feathers. It leads people to anthropomorphise, and you see that in discussions of whether "AI is lying to you" (it's not, it can't lie, it's just wrong), or "bullshitting". Even the preferred term of "hallucinating" is implying a level of agency that just isn't there.
I wonder if we need to collectively discuss more appropriate terminology. I prefer to call it 'generative computing' where possible, but always have to explain what that is. Same with using the initialism LLM. Using 'AI' often comes with the subtext that the software under discussion is, or is about to become, sentient, and that then brings with it a set of assumptions about intention, rights and personality, and the creation of parasocial relationships, that are deeply troubling.
Blessed be those taking an intelligent approach framed in an international context (or any context that doesn't bipolarise bots between skynet and BS hype-scent). It would be interesting to extend your analysis into an East vs West treatment.
My two other main concerns beyond Americani[s|z]ation of IP are:
1) Filtering: Yes, chatbots will turn most writing into 'Instagram for words'. Worse, these bots will even be used to provide the initial prompts for creating awkward, statistically relevant, yet emotionally empty text. But, worst of all, these prompts will be pre-filtered into 'socially corrected' memes by those who see their job as protecting free speech +/or social harmony whilst simultaneously denying +/or inflicting both.
2) Revisionism:
Rewriting text to ensure eternal cultural validity will become a culture-washing role, primarily for institutionally owned publishers, but also socially acceding grafted writers, politicians will hardly hinder the erasure of culpable error. Not only feeding the unleashed AI beasts, we'll also help prune and burn their uncomfortable parts. Sifting through the digital ash from bot-driven revision, historians will never divine the truth from the Corpus of history. We'll never be allowed to know it. We'll all unwittingly become willing Winstons.
Maybe information wants to be free. Work always has a cost -- that's true in the marketplace and it's true in physics. My view: a bit of software that incorporates my work in its own products has to pay for the right.
Anyway -- great piece.
Many thanks for being at the forefront of intellectual property in the neural network age. While I don't believe for a second that the current generation of AIs—ChatGPT, Bing or Google's Bard—have anything resembling actual intelligence, I do see that they copy existing art.
I can't even say 'shamelessly' as these neural networks are devoid of emotions like 'shame'. Their proprietors may have a semblance of integrity hidden somewhere deep inside, but who am I fooling? The moment those neural networks generate big bucks, the plagiarised artists are on their own, without the massive amount of lawyers the likes of Google, Microsoft and Meta have by the bucketload.
It makes me extremely reluctant to post even small parts of a WiP (work in progress) online, and wonder if I should publish on Google Play, as this means your work is automatically put on Google Books, and is subsequently training material for their Google Bard neural network.
It's a great point that you can't predict the next Rembrandt painting by averaging all that came before. One might even argue that his paintings diverged increasingly from being 'average Rembrandt' over time - and became more essentially Rembrandt (which aligns with the German IP approach). My go-to counter to the hype is, sure, your machine model can generate a cubist-look image - but can it invent cubism?
As always Nick, it's both your posts and the comments on here that are so valuable. Re-read EM Foster's "The Duty of Society to the Artist" essay this morning, and 'generative computing' seems to be the 'centralising and disappearing' of the artist that Forster feared.
AI as a derivative work is fascinating and rings... true. The foundation and meaning behind the different copyright and IP laws you shared is fascinating, and I would love to learn more and about more of them, what are or were other traditions of that? Have you read this? https://www.noemamag.com/ai-is-life
Yes we need UBI or whatever the global version of nationalization is or ownership by all of the creators, which as this noama article argues could be everyone, ever.