MacDonaldStrand’s latest project – a self-published photobook ‘The Automata Chronicles: The Age of Ghostwriters’ – is a contemporary AI-generated fable telling the rise of artificially intelligent machines, and in so doing replacing one of our most “cherished abilities: the art of writing.”
In ceding writing to machines, the line between author and reader no longer exists, nor do celebrated poets, authors, or playwrights. Instead, there is endless generated content, served by algorithms to satisfy user preferences. Interactions now occur solely with machines and the art of conversation and connection is lost. In the final act, humans fight back – reverse engineering AI algorithms to regain their writing skills and knowledge – but much has already been lost.
Matthew, Clare, and Gordon convened by email to discuss the book and the wider implications of AI-generated images.
MW: Perhaps we could start by describing the intent behind the book, the story, format, and the response to the work?
MDS: The intent was simply to engage with AI text and image generators, and to ask them a simple but searching question. When we started we had no idea that it would be worthwhile, only that we would see how this simple question would be answered.
MW: Could you describe the background/context in which this work was produced?
MDS: For a couple of decades now, we have all been pouring our thoughts, ideas, memories and records into a machine, both as text and images. Our home lives, work lives, secret lives and love lives are all stored there: our likes, dislikes; our financial histories, our debts and our purchases all build up a picture of who we are. We all routinely accept this when we check ‘accept’ on the terms and conditions statements that pop up on our screens. We agreed and continued to agree, in order to use online services and social media platforms. We were offered a simple choice – join in, feed in, or slip out of the conversation and ‘know’ less than others, have less control than others, be less in the world than others. In retrospect, with social media at least, it wasn’t much of a choice.
More recently, this legacy of data, the texts and images we have amassed in the machine, have begun to be harvested by generative AI. Every great work of literature, every historic speech, every note to a loved one or gripe about politics; everything written and shared online. Images also become part of this information pool – from holiday snaps, baby photos, cat photos wedding photos, to pornographic imagery, or photographs of devastating war, destruction and suffering.
MW: How were the images and text produced? Was the output from the AI generators edited and refined? Was there a learning, repeating, transforming process?
MDS: The starting point for this project was the simple prompt to an online generative AI text programme, asking it to imagine ‘A dystopian world, where AI is trusted to write books.’ The resulting text, produced in seconds by the AI machine, is the story of a world in which humans are pushed aside, following a lazy capitulation to GPT. In this text we see the demise of the humans, followed by an enlightenment and subsequent revolution against the machines, and of course triumph. It follows the understandable and often-explored theme of peoples fight for freedom from technology – a fight for the ‘human spirit and soul’ over cold metal and algorithm. The story could have been written in Hollywood for a movie starring Arnold Schwarzenegger or Tom Cruise. But the text also seems to reflect an uncertainty within the AI itself – as if it understands the problems of this move towards self-aware machines. It is like the AI is having an existential crisis.
The chapter headings of the generated story were passed from the AI text generator to an AI image generator, so that the story, which has the rhythm and chapter lengths of a children’s story book, could be illustrated. Within the images produced it is easy to see the words and their impact as prompts. Again, like a children’s book, they are extremely literal. But hidden in these images are the fragments of our digital legacy.
There was no intervention in either the text or the images by us, other than our selection of which response we wanted to accept from the machine.
MW: The images in the book are striking. Laying on the page in sharp, high-contrast, black-and-white form they evoke Lewis Hine’s work in the 1920s US industrial landscape, placing man alongside and perhaps part of machine such as in the classic steam pump photo. Further comparisons could be made with films such as Metropolis with man as the machine, and machine as man, as well as photographers such as Marcel Gautherot in his architectural work and Cindy Sherman’s portraits. Was there an intent to convey these specific styles through the prompts or post-processing? What does it say that the styles are clearly recognisable and AI is often personified?
MDS: Initially, it is simple to spot the influence of 19th and 20th century photographic history, from Man Ray to Lewis Hine, as well as classic movie imaginations of self-aware automaton like Metropolis or Blade Runner. But there are also the deeper-mined fragments of people – their lips, eyes and hands. These are the fragments of us and of our loved ones, and the legacy of our unknowing work for the machines. These images of ‘people’ are us, but not; human and machine, no more ‘real’ than the machines imagined images of itself.
These easy to recognise styles and tropes point to the infancy of the learning process of AI. Like a human infant, it seems to us to be trying to please the adults. It learns and regurgitates – sometimes coherently and sometimes slightly muddled, but always looking for what may answer a question using the material it has available.
MW: In a recent test case, the US government ruled that an image which Matthew Allen created by an AI image generator, and won the Colorado State Art Fair’s main prize, cannot be copyrighted as it was predominantly made by a machine, rather than a human. He could, however, copyright elements of the image that he had altered. In your view, who should own the copyright of AI-produced works and how do rulings like this affect your own work and that of others who are considering making images using AI?
MDS: It poses a potentially disastrous scenario, where a machine could own the copyright, or worse, the tech billionaire who owns the machine. The machine is afterall ‘owned’ by a human. Given this line of reason, the camera would hold copyright on the photographs it made, or the brush on the paintings it touched – or the people who manufactured them. But for us, the concept is the artwork, not the product – the product is just an illustration or proof of the concept. We are pretty sure that it will not be the machine challenging us on this.
MW: Authenticity has long been prized in photography with traditional processes held sacrosanct by some while the wider industry moves forward with ever wider-ranging technological advancements, folding these into a wider array of image styles and forms. How do you see AI-generated images affecting the photography industry, the work people make and the projects people take on?
MDS: As we mentioned, photography has always been made by machines. We own analogue and digital equipment to make photographic images (if we ever do decide to make photographs), and try to use the appropriate equipment for the project we are working on. For us, it is a crazy idea to disregard any method of making, any source material or any potential outcome. We will continue to engage with technological changes and consider their uses, but will also not become evangelical about it.
Of course the effect of new digital possibilities will change the way images are made in new and interesting ways, and this might alienate many who feel that the ‘industry’ is being changed too much. For our part, we have never felt part of an ‘industry’, as the idea that an interest in cameras or prints should link people’s practice seems to be less useful than ideas or conceptual interests linking artists and makers, whatever their mode of practice.
MW: Do we need a separate term such as Promptography to differentiate between computer and human-made images? Should AI images be allowed to compete in art and photography competitions or be published on large media platforms?
MDS: One could invent any term, and a lot of people seem to be wanting to define what we should call these AI-produced works. But they are still just images, no matter how much language we invent for them. Obviously, artists and makers should be honest about the method of production of the images that they are making. But the fact that any artform has to rely on prizes is a bit demeaning in the first place, regardless of the production method. And that a competition jury could compare a photograph of a warzone to a photograph of a cat and decide which is the best image, is to miss the point of both works, whether made by a conventional camera or by using a computer. It feels odd that the conversation is skewed so far towards the technology used, rather than the content or concept of a work.
MW: We see companies such as Leica incorporate security chips in their latest M11-P camera to verify authorship of the images produced from the camera and any edits that are made subsequent to the initial capture. This could be helpful for those who purchase images or allot awards in competitions. I would also value this to show ownership of the images I’ve made. Do you think we’ll see more affordable cameras as well as digital art devices adding this into their products and a chain of software and registries that support ownership and do you think it’s necessary?
MDS: We would imagine that camera companies, like all businesses driven by a profit motive, will do what sells the most product and makes their owners and shareholders the most money. We doubt that they are driven by the image rights of their customers outside of it’s overlap with these concerns. But neither of us are knowledgeable enough about camera development and technology to second-guess beyond this.
How should social media and larger media platforms use this type of verification to label images on their medium and equally what authorship labels should we apply to AI-generated content?
MDS: Again, we are all pretty much slaves to social media companies, and we doubt that they will do much unless share value is threatened.
MW: How does this work sit with other projects that you have done individually or together? Does it change how you will work from now on? How does it change how you view the world?
MDS: As MacDonaldStrand we like to make work which reflects some aspect of the use of images in society. This work is part of that practice. It doesn’t change how we may make work in the future any more than previous projects using heavily edited press imagery, dot-to-dot renderings of notable photographs or sweary publications about the photobook market have influenced this project. We make each work as a stand-alone statement. And we view the world with a mixture of cynicism and awe – the same as we did before AI technology became part of everyone’s conversation.
MW: Prior to our conversation, I wondered what would happen if AI tools were prompted to create an art project without a specific theme. What would they choose? When using Google’s Bard platform, all responses came back with some reference to consciousness or the human condition, though all with a connection to technology and often AI itself. Perhaps there is a ghost in the machine after all. In your experience exploring AI do you feel the emphasis on the human condition that we have seen since photographers such as Hine endures?
MDS: AI, in our experience here, seems to be very good at reflecting shared or regularly-occurring concerns. In this project, the AI was asked a very simple question through the prompt. What it returned was far beyond the scope of this question, and was incredibly poignant to contemporary concerns around AI. This is not accidental or magical – it is what the machine has been asked to do through it’s algorithm. The more we feed it these concerns or aspirations, the more strongly it will reflect them. In this project the AI-produced images also seem to rely heavily on the American archives and copyright-free material as the reference pool. The future tipping-point will be when the machine is drawing largely on previously AI-produced images and text as the source material. Maybe we will make another AI project at that point, to see what has changed.
https://www.macdonaldstrand.co.uk/store/p/the-automata-chronicles-the-age-of-the-ghostwriters
MacDonaldStrand’s latest project – a self-published photobook ‘The Automata Chronicles: The Age of Ghostwriters’ – is a contemporary AI-generated fable telling the rise of artificially intelligent machines, and in so doing replacing one of our most “cherished abilities: the art of writing.”
In ceding writing to machines, the line between author and reader no longer exists, nor do celebrated poets, authors, or playwrights. Instead, there is endless generated content, served by algorithms to satisfy user preferences. Interactions now occur solely with machines and the art of conversation and connection is lost. In the final act, humans fight back – reverse engineering AI algorithms to regain their writing skills and knowledge – but much has already been lost.
Matthew, Clare, and Gordon convened by email to discuss the book and the wider implications of AI-generated images.
MW: Perhaps we could start by describing the intent behind the book, the story, format, and the response to the work?
MDS: The intent was simply to engage with AI text and image generators, and to ask them a simple but searching question. When we started we had no idea that it would be worthwhile, only that we would see how this simple question would be answered.
MW: Could you describe the background/context in which this work was produced?
MDS: For a couple of decades now, we have all been pouring our thoughts, ideas, memories and records into a machine, both as text and images. Our home lives, work lives, secret lives and love lives are all stored there: our likes, dislikes; our financial histories, our debts and our purchases all build up a picture of who we are. We all routinely accept this when we check ‘accept’ on the terms and conditions statements that pop up on our screens. We agreed and continued to agree, in order to use online services and social media platforms. We were offered a simple choice – join in, feed in, or slip out of the conversation and ‘know’ less than others, have less control than others, be less in the world than others. In retrospect, with social media at least, it wasn’t much of a choice.
More recently, this legacy of data, the texts and images we have amassed in the machine, have begun to be harvested by generative AI. Every great work of literature, every historic speech, every note to a loved one or gripe about politics; everything written and shared online. Images also become part of this information pool – from holiday snaps, baby photos, cat photos wedding photos, to pornographic imagery, or photographs of devastating war, destruction and suffering.
MW: How were the images and text produced? Was the output from the AI generators edited and refined? Was there a learning, repeating, transforming process?
MDS: The starting point for this project was the simple prompt to an online generative AI text programme, asking it to imagine ‘A dystopian world, where AI is trusted to write books.’ The resulting text, produced in seconds by the AI machine, is the story of a world in which humans are pushed aside, following a lazy capitulation to GPT. In this text we see the demise of the humans, followed by an enlightenment and subsequent revolution against the machines, and of course triumph. It follows the understandable and often-explored theme of peoples fight for freedom from technology – a fight for the ‘human spirit and soul’ over cold metal and algorithm. The story could have been written in Hollywood for a movie starring Arnold Schwarzenegger or Tom Cruise. But the text also seems to reflect an uncertainty within the AI itself – as if it understands the problems of this move towards self-aware machines. It is like the AI is having an existential crisis.
The chapter headings of the generated story were passed from the AI text generator to an AI image generator, so that the story, which has the rhythm and chapter lengths of a children’s story book, could be illustrated. Within the images produced it is easy to see the words and their impact as prompts. Again, like a children’s book, they are extremely literal. But hidden in these images are the fragments of our digital legacy.
There was no intervention in either the text or the images by us, other than our selection of which response we wanted to accept from the machine.
MW: The images in the book are striking. Laying on the page in sharp, high-contrast, black-and-white form they evoke Lewis Hine’s work in the 1920s US industrial landscape, placing man alongside and perhaps part of machine such as in the classic steam pump photo. Further comparisons could be made with films such as Metropolis with man as the machine, and machine as man, as well as photographers such as Marcel Gautherot in his architectural work and Cindy Sherman’s portraits. Was there an intent to convey these specific styles through the prompts or post-processing? What does it say that the styles are clearly recognisable and AI is often personified?
MDS: Initially, it is simple to spot the influence of 19th and 20th century photographic history, from Man Ray to Lewis Hine, as well as classic movie imaginations of self-aware automaton like Metropolis or Blade Runner. But there are also the deeper-mined fragments of people – their lips, eyes and hands. These are the fragments of us and of our loved ones, and the legacy of our unknowing work for the machines. These images of ‘people’ are us, but not; human and machine, no more ‘real’ than the machines imagined images of itself.
These easy to recognise styles and tropes point to the infancy of the learning process of AI. Like a human infant, it seems to us to be trying to please the adults. It learns and regurgitates – sometimes coherently and sometimes slightly muddled, but always looking for what may answer a question using the material it has available.
MW: In a recent test case, the US government ruled that an image which Matthew Allen created by an AI image generator, and won the Colorado State Art Fair’s main prize, cannot be copyrighted as it was predominantly made by a machine, rather than a human. He could, however, copyright elements of the image that he had altered. In your view, who should own the copyright of AI-produced works and how do rulings like this affect your own work and that of others who are considering making images using AI?
MDS: It poses a potentially disastrous scenario, where a machine could own the copyright, or worse, the tech billionaire who owns the machine. The machine is afterall ‘owned’ by a human. Given this line of reason, the camera would hold copyright on the photographs it made, or the brush on the paintings it touched – or the people who manufactured them. But for us, the concept is the artwork, not the product – the product is just an illustration or proof of the concept. We are pretty sure that it will not be the machine challenging us on this.
MW: Authenticity has long been prized in photography with traditional processes held sacrosanct by some while the wider industry moves forward with ever wider-ranging technological advancements, folding these into a wider array of image styles and forms. How do you see AI-generated images affecting the photography industry, the work people make and the projects people take on?
MDS: As we mentioned, photography has always been made by machines. We own analogue and digital equipment to make photographic images (if we ever do decide to make photographs), and try to use the appropriate equipment for the project we are working on. For us, it is a crazy idea to disregard any method of making, any source material or any potential outcome. We will continue to engage with technological changes and consider their uses, but will also not become evangelical about it.
Of course the effect of new digital possibilities will change the way images are made in new and interesting ways, and this might alienate many who feel that the ‘industry’ is being changed too much. For our part, we have never felt part of an ‘industry’, as the idea that an interest in cameras or prints should link people’s practice seems to be less useful than ideas or conceptual interests linking artists and makers, whatever their mode of practice.
MW: Do we need a separate term such as Promptography to differentiate between computer and human-made images? Should AI images be allowed to compete in art and photography competitions or be published on large media platforms?
MDS: One could invent any term, and a lot of people seem to be wanting to define what we should call these AI-produced works. But they are still just images, no matter how much language we invent for them. Obviously, artists and makers should be honest about the method of production of the images that they are making. But the fact that any artform has to rely on prizes is a bit demeaning in the first place, regardless of the production method. And that a competition jury could compare a photograph of a warzone to a photograph of a cat and decide which is the best image, is to miss the point of both works, whether made by a conventional camera or by using a computer. It feels odd that the conversation is skewed so far towards the technology used, rather than the content or concept of a work.
MW: We see companies such as Leica incorporate security chips in their latest M11-P camera to verify authorship of the images produced from the camera and any edits that are made subsequent to the initial capture. This could be helpful for those who purchase images or allot awards in competitions. I would also value this to show ownership of the images I’ve made. Do you think we’ll see more affordable cameras as well as digital art devices adding this into their products and a chain of software and registries that support ownership and do you think it’s necessary?
MDS: We would imagine that camera companies, like all businesses driven by a profit motive, will do what sells the most product and makes their owners and shareholders the most money. We doubt that they are driven by the image rights of their customers outside of it’s overlap with these concerns. But neither of us are knowledgeable enough about camera development and technology to second-guess beyond this.
How should social media and larger media platforms use this type of verification to label images on their medium and equally what authorship labels should we apply to AI-generated content?
MDS: Again, we are all pretty much slaves to social media companies, and we doubt that they will do much unless share value is threatened.
MW: How does this work sit with other projects that you have done individually or together? Does it change how you will work from now on? How does it change how you view the world?
MDS: As MacDonaldStrand we like to make work which reflects some aspect of the use of images in society. This work is part of that practice. It doesn’t change how we may make work in the future any more than previous projects using heavily edited press imagery, dot-to-dot renderings of notable photographs or sweary publications about the photobook market have influenced this project. We make each work as a stand-alone statement. And we view the world with a mixture of cynicism and awe – the same as we did before AI technology became part of everyone’s conversation.
MW: Prior to our conversation, I wondered what would happen if AI tools were prompted to create an art project without a specific theme. What would they choose? When using Google’s Bard platform, all responses came back with some reference to consciousness or the human condition, though all with a connection to technology and often AI itself. Perhaps there is a ghost in the machine after all. In your experience exploring AI do you feel the emphasis on the human condition that we have seen since photographers such as Hine endures?
MDS: AI, in our experience here, seems to be very good at reflecting shared or regularly-occurring concerns. In this project, the AI was asked a very simple question through the prompt. What it returned was far beyond the scope of this question, and was incredibly poignant to contemporary concerns around AI. This is not accidental or magical – it is what the machine has been asked to do through it’s algorithm. The more we feed it these concerns or aspirations, the more strongly it will reflect them. In this project the AI-produced images also seem to rely heavily on the American archives and copyright-free material as the reference pool. The future tipping-point will be when the machine is drawing largely on previously AI-produced images and text as the source material. Maybe we will make another AI project at that point, to see what has changed.
https://www.macdonaldstrand.co.uk/store/p/the-automata-chronicles-the-age-of-the-ghostwriters
