You carefully share detail on the prompts used in your AI generated image; can you also credit the artist’s, whose work, without consent or compensation, trained the database needed to generate that image?
Hi Lezley. I hear what you are saying here, and I like the fact that you care about the rights of artists. I am intensely aware of these issues myself: at least two of my novels were in the enormous trove of pirated books which formed part of the notorious Books3 data set, part of a larger dataset called The Pile, which has been used by most of the main AI companies to train their Large Language Models (LLMs). So anytime you use Chat-GPT, some of its response was very likely trained on my copyrighted writings, without my permission or consent. (Would I have granted it if asked? That's a whole other question.; they didn't ask, they took without asking.) I also recently wrote about this issue for The Bookseller, the magazine of the Society of Authors (in a piece called AI and the Future of Authorship).
If it helps, I deliberately asked for an 18th century woodcut, so the training data should minimally infringe the copyrights of living artists.
But... I also think that there is more to life than the maximisation of copyright protection by artists. As you may know, I gave away the copyright to the ending to Minecraft (the End Poem) last year, and put it in the public domain under a Creative Commons licence, because I think some art (maybe most art... maybe all art) shouldn't be owned by anyone. I would like different models to apply, that allowed far more artists to survive and make art; that, ideally, allowed us all the space to express ourselves through art without starving or becoming homeless. (And I speak as an artist who has indeed been evicted while working hard on his art, because that art didn't generate enough money to live on.) How do we get there from here? I don't know, but I am working on it. Talking to others. Trying to find new economic models for the arts.
Meanwhile, I am fascinated by LLMs, and by generative art, and I feel, as an artist who isn't charging for his work here, I have an artistic right (a morally dubious one, sure, as is the case with so much of art!) to play with these new tools, which is separate from the legal and copyright issues.
So, one way of looking at this is that I am an artist using a new tool to play with the collective unconscious of art. This is a glorious thing! Are there difficult moral issues around this? Sure! Am I a saint? No! Is the commercial exploitation of broke artists by incredibly rich corporations kind of disgusting? Yep! But is it also leading to something new under the sun that is kind of thrilling? Also yep!
I realise I am wading thigh deep in an ethical swamp here, but if all the artists stay out of the swamp, and leave it entirely to the corporations... will that make for a better world? Or will it just make us feel good about ourselves – high above the struggle, atop Mount Morality – without in fact making anything better.
Anyway, I would be happy to credit all the artists who ever lived, and whose work was scraped in the training data for DALLE-E. I owe you, my brothers and sisters in art. We all owe you more than we will ever pay you; more than you were ever paid.
I appreciate your thorough reply and also don't have any answers or solutions for the current situation. I do know that I would feel differently if the scraping happened to create a resource that was open and free for all to share and contribute and benefit from. That’s not what happened. If you feel entitled to use an AI generator because you provide art to the public for free… how do you feel about someone who uses the same tools but has contributed no art of their own? Does that change your perception of their right to access?
"How do you feel about someone who uses the same tools but has contributed no art of their own? Does that change your perception of their right to access?"
I am very non-judgemental in this area. I like it when anyone at all tries to make art, even if they use morally dubious AI tools to do so. (Clearly prompting an AI isn't the same as painting an oil painting. But most people are not painting an oil painting; that is not what AI is displacing.) I am very accepting of the moral messiness of art and art-making and art ownership. But I am also very accepting of the fact that there are other ways to look at this. I'm not saying you are wrong and I am right: I am saying we are different and see these things differently, and that is fine. That is necessary.
If it helps to know: A couple of years back, as rumours of new, powerful generative AI tools started to leak from the labs, I was asked to join an IEEE committee to develop standards for AI in the arts. (IEEE is the Institute of Electrical and Electronics Engineers, and they set global standards for all kinds of technologies.) It was a good, hardworking committee of excellent people, most of them artists. And after a few months of participation, I resigned and walked away from it because I felt my level of uncertainty about any recommendations I could make was far too high. I literally could not tell if my recommendations were good or bad; would have good effects, or bad ones. I had thought about this stuff for MONTHS, in a very focused way, and at the end of it my feeling was this: AI is definitely coming. It will definitely transform the world. It will definitely transform the arts. But I have no idea in what ways. I have no idea what is the best course of action for humanity in relation to AI. I have no idea what direction we should try to nudge it in. I have no idea what the ACTUAL effects of any proposed nudges, regulations, laws, or standards would actually be. Any time you try to think through the consequences of an action, there is an immediate cascade of second-order and third-order consequences that blur out any clear vision of the future.
I have never felt so much uncertainty as I do around the subject of AI. And so I play around with it, I interrogate it, I explore it. And I try to have the humility to not impose my shifting, uncertain opinions on others, because I don't think any of us know – any of us CAN know – what will happen next. It's potentially a breakthrough as important as the first tool use by humans; the woven grass carrier bag, the chipped flint ax. And tools since then have done a lot of good and a lot of harm. But, having discovered tools, you can't undiscover them. Groups can give them up, can renounce them, can try to ban them., But others, elsewhere, will go on exploring their use, and those groups who have explored the new tools will soon be far more powerful than those who have renounced the tools. So this is going to happen. Hang on tight...
It's an interesting quandary. On one hand, I can easily see how someone who has written / drawn / produced art would have more of a 'title' as it were to utilize such tools.
On the other, where would you set the threshold? Is one short story enough? One novel? One painting? What is the 'minimum' as it were?
The other thing is...even if they have not, if they are fools with no taste, what they produce will be similarly dumb - as we see so often on social media with 'Oh I talked to Dall-E and-' because what matters, on some level, is your taste & vision. Not just 'can you ask the LLM' but 'Are you capable of gleaning what is worth showing others versus what is crap' from what it wells up, because its very easy to make them produce crap.
If artists weren't exploited by corporations, if we had a more just & equitable world where the amazing bounty of human ingenuity was shared more widely rather than hoarded so billionaires can have biggest yacht competitions, I think this would be a non-issue.
The problem here isn't the training - the problem is artists being terrified for their livelihoods (justifiably, unfortunately) because the people in charge of companies are idiots without taste who are fine with regurgitated papp because they are idiots without taste and so are okay with Digital Nutri-Slop.
Those of us who crave real artistic meals are the ones all enraged - but the same cultural vultures doing this are the ones who've ruined so much of the rest of the world. Bezos's latest yacht cost 500 million USD. The US GDP per capita is about 70,000 USD a year.
Or, in other words, 178.5 people's entire productive lifetimes of 40 years went into building this yacht. Imagine would good could have done had that time been dedicated to helping people, rather than building a giant dick measuring toy.
That's the thing worth getting angry over, imo, and the AI art fight is way to make creatives fight amongst ourselves rather than focus on the real enemy - especially as I agree with Julian (if I am parsing you correctly, please correct if wrong) that the coming AI flowering is probably inevitable, unless the NYT wins its lawsuit.
In short, there's a question few are considering, which is 'Given how expensive LLMs are to run, are they able to be profitable?' because if there arent enough people willing to pay once the M$ subsidy dries up a lot of this is going to collapse.
lol “digits nutri-slop “ There is some push back in the form of Glaze and other digital filters that prevent further scraping and can actually poison the dataset.
I am entirely in favor of Glaze and other tools - it fixes the consent issue and I am kind of fine with those models being poisoned. Also, given Chat GPT is already engorging its own slop I kind of hope it self-poisons as well. It would be hilariously ironic in the best of ways.
Yes, Glaze is an interesting first response by the artistic immune system. An early example of the kind of non-obvious consequence of AI that makes predicting the future so difficult. (Impossible!) It is going to be SO INTERESTING, as the entire digital arts ecosystem starts to respond to AI disruption. There will be extinctions, and new species...
You carefully share detail on the prompts used in your AI generated image; can you also credit the artist’s, whose work, without consent or compensation, trained the database needed to generate that image?
Hi Lezley. I hear what you are saying here, and I like the fact that you care about the rights of artists. I am intensely aware of these issues myself: at least two of my novels were in the enormous trove of pirated books which formed part of the notorious Books3 data set, part of a larger dataset called The Pile, which has been used by most of the main AI companies to train their Large Language Models (LLMs). So anytime you use Chat-GPT, some of its response was very likely trained on my copyrighted writings, without my permission or consent. (Would I have granted it if asked? That's a whole other question.; they didn't ask, they took without asking.) I also recently wrote about this issue for The Bookseller, the magazine of the Society of Authors (in a piece called AI and the Future of Authorship).
If it helps, I deliberately asked for an 18th century woodcut, so the training data should minimally infringe the copyrights of living artists.
But... I also think that there is more to life than the maximisation of copyright protection by artists. As you may know, I gave away the copyright to the ending to Minecraft (the End Poem) last year, and put it in the public domain under a Creative Commons licence, because I think some art (maybe most art... maybe all art) shouldn't be owned by anyone. I would like different models to apply, that allowed far more artists to survive and make art; that, ideally, allowed us all the space to express ourselves through art without starving or becoming homeless. (And I speak as an artist who has indeed been evicted while working hard on his art, because that art didn't generate enough money to live on.) How do we get there from here? I don't know, but I am working on it. Talking to others. Trying to find new economic models for the arts.
Meanwhile, I am fascinated by LLMs, and by generative art, and I feel, as an artist who isn't charging for his work here, I have an artistic right (a morally dubious one, sure, as is the case with so much of art!) to play with these new tools, which is separate from the legal and copyright issues.
So, one way of looking at this is that I am an artist using a new tool to play with the collective unconscious of art. This is a glorious thing! Are there difficult moral issues around this? Sure! Am I a saint? No! Is the commercial exploitation of broke artists by incredibly rich corporations kind of disgusting? Yep! But is it also leading to something new under the sun that is kind of thrilling? Also yep!
I realise I am wading thigh deep in an ethical swamp here, but if all the artists stay out of the swamp, and leave it entirely to the corporations... will that make for a better world? Or will it just make us feel good about ourselves – high above the struggle, atop Mount Morality – without in fact making anything better.
Anyway, I would be happy to credit all the artists who ever lived, and whose work was scraped in the training data for DALLE-E. I owe you, my brothers and sisters in art. We all owe you more than we will ever pay you; more than you were ever paid.
This wonderfully encapsulates some of my own feelings on AI generated art. Will be quoting & linking to this!
Hey, thanks!
I appreciate your thorough reply and also don't have any answers or solutions for the current situation. I do know that I would feel differently if the scraping happened to create a resource that was open and free for all to share and contribute and benefit from. That’s not what happened. If you feel entitled to use an AI generator because you provide art to the public for free… how do you feel about someone who uses the same tools but has contributed no art of their own? Does that change your perception of their right to access?
"How do you feel about someone who uses the same tools but has contributed no art of their own? Does that change your perception of their right to access?"
I am very non-judgemental in this area. I like it when anyone at all tries to make art, even if they use morally dubious AI tools to do so. (Clearly prompting an AI isn't the same as painting an oil painting. But most people are not painting an oil painting; that is not what AI is displacing.) I am very accepting of the moral messiness of art and art-making and art ownership. But I am also very accepting of the fact that there are other ways to look at this. I'm not saying you are wrong and I am right: I am saying we are different and see these things differently, and that is fine. That is necessary.
If it helps to know: A couple of years back, as rumours of new, powerful generative AI tools started to leak from the labs, I was asked to join an IEEE committee to develop standards for AI in the arts. (IEEE is the Institute of Electrical and Electronics Engineers, and they set global standards for all kinds of technologies.) It was a good, hardworking committee of excellent people, most of them artists. And after a few months of participation, I resigned and walked away from it because I felt my level of uncertainty about any recommendations I could make was far too high. I literally could not tell if my recommendations were good or bad; would have good effects, or bad ones. I had thought about this stuff for MONTHS, in a very focused way, and at the end of it my feeling was this: AI is definitely coming. It will definitely transform the world. It will definitely transform the arts. But I have no idea in what ways. I have no idea what is the best course of action for humanity in relation to AI. I have no idea what direction we should try to nudge it in. I have no idea what the ACTUAL effects of any proposed nudges, regulations, laws, or standards would actually be. Any time you try to think through the consequences of an action, there is an immediate cascade of second-order and third-order consequences that blur out any clear vision of the future.
I have never felt so much uncertainty as I do around the subject of AI. And so I play around with it, I interrogate it, I explore it. And I try to have the humility to not impose my shifting, uncertain opinions on others, because I don't think any of us know – any of us CAN know – what will happen next. It's potentially a breakthrough as important as the first tool use by humans; the woven grass carrier bag, the chipped flint ax. And tools since then have done a lot of good and a lot of harm. But, having discovered tools, you can't undiscover them. Groups can give them up, can renounce them, can try to ban them., But others, elsewhere, will go on exploring their use, and those groups who have explored the new tools will soon be far more powerful than those who have renounced the tools. So this is going to happen. Hang on tight...
It's an interesting quandary. On one hand, I can easily see how someone who has written / drawn / produced art would have more of a 'title' as it were to utilize such tools.
On the other, where would you set the threshold? Is one short story enough? One novel? One painting? What is the 'minimum' as it were?
The other thing is...even if they have not, if they are fools with no taste, what they produce will be similarly dumb - as we see so often on social media with 'Oh I talked to Dall-E and-' because what matters, on some level, is your taste & vision. Not just 'can you ask the LLM' but 'Are you capable of gleaning what is worth showing others versus what is crap' from what it wells up, because its very easy to make them produce crap.
If artists weren't exploited by corporations, if we had a more just & equitable world where the amazing bounty of human ingenuity was shared more widely rather than hoarded so billionaires can have biggest yacht competitions, I think this would be a non-issue.
The problem here isn't the training - the problem is artists being terrified for their livelihoods (justifiably, unfortunately) because the people in charge of companies are idiots without taste who are fine with regurgitated papp because they are idiots without taste and so are okay with Digital Nutri-Slop.
Those of us who crave real artistic meals are the ones all enraged - but the same cultural vultures doing this are the ones who've ruined so much of the rest of the world. Bezos's latest yacht cost 500 million USD. The US GDP per capita is about 70,000 USD a year.
Or, in other words, 178.5 people's entire productive lifetimes of 40 years went into building this yacht. Imagine would good could have done had that time been dedicated to helping people, rather than building a giant dick measuring toy.
That's the thing worth getting angry over, imo, and the AI art fight is way to make creatives fight amongst ourselves rather than focus on the real enemy - especially as I agree with Julian (if I am parsing you correctly, please correct if wrong) that the coming AI flowering is probably inevitable, unless the NYT wins its lawsuit.
Although Cory Doctorow has some great thoughts on this over on his blog - https://pluralistic.net/2023/12/19/bubblenomics/
In short, there's a question few are considering, which is 'Given how expensive LLMs are to run, are they able to be profitable?' because if there arent enough people willing to pay once the M$ subsidy dries up a lot of this is going to collapse.
lol “digits nutri-slop “ There is some push back in the form of Glaze and other digital filters that prevent further scraping and can actually poison the dataset.
I am entirely in favor of Glaze and other tools - it fixes the consent issue and I am kind of fine with those models being poisoned. Also, given Chat GPT is already engorging its own slop I kind of hope it self-poisons as well. It would be hilariously ironic in the best of ways.
Yes, Glaze is an interesting first response by the artistic immune system. An early example of the kind of non-obvious consequence of AI that makes predicting the future so difficult. (Impossible!) It is going to be SO INTERESTING, as the entire digital arts ecosystem starts to respond to AI disruption. There will be extinctions, and new species...