INTERVIEW WITH K ALLADO-MCDOWELL
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
K Allado-McDowell is a writer, artist, musician, and a pioneering figure in the field of AI literature. Since 2020, they have authored several books with GPT-3, co-edited and contributed to multiple anthologies, and regularly publish essays on art, AI, and ecology.
K established the Artists + Machine Intelligence program at Google in 2015. They created the neuro-opera Song of the Ambassadors, which previewed at Lincoln Center in 2022. Their first solo exhibition,
The Known Lost, was presented at Swiss Institute in NYC in 2025.
INTERVIEW WITH K ALLADO-MCDOWELL
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
K Allado-McDowell is a writer, artist, musician, and a pioneering figure in the field of AI literature. Since 2020, they have authored several books with GPT-3, co-edited and contributed to multiple anthologies, and regularly publish essays on art, AI, and ecology.
K established the Artists + Machine Intelligence program at Google in 2015. They created the neuro-opera Song of the Ambassadors, which previewed at Lincoln Center in 2022. Their first solo exhibition,
The Known Lost, was presented at Swiss Institute in NYC in 2025.
INTERVIEW WITH K ALLADO-MCDOWELL
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
K Allado-McDowell is a writer, artist, musician, and a pioneering figure in the field of AI literature. Since 2020, they have authored several books with GPT-3, co-edited and contributed to multiple anthologies, and regularly publish essays on art, AI, and ecology.
K established the Artists + Machine Intelligence program at Google in 2015. They created the neuro-opera Song of the Ambassadors, which previewed at Lincoln Center in 2022. Their first solo exhibition,
The Known Lost, was presented at Swiss Institute in NYC in 2025.
INTERVIEW WITH K ALLADO-MCDOWELL
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
JF: What has been your process and experience working with AI? Is it just a tool or do you relate to it as a colleague and collaborator?
K A-M: I was interested in constructing situations that were kind of open-ended, like mashing two concepts together and seeing what came out. If, for example, you put in a really opinionated diatribe about something, it will just give you more of that. So it is less about it as an objective system that has its own specific opinion that is smarter or truer than ours; it’s more like an oracle or something you speak into that reflects something back to you that may be enigmatic and require interpretation on your part.
Generally, you want a tool to be predictable and precise, to respond in a specific way. Neural networks are good at producing very specific results, but what we find interesting about them is when they create something we weren’t expecting. An artificially-intelligent neural structure is highly flexible and non-linear, similarly the brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Thinking of it in terms of it as an entity unto itself, or a collaborator or a partner, is quite difficult because it’s a very shape-shifting kind of thing. Learning its patterns and understanding how to get what you want out of it is part of it, but so is understanding how it pushes back..
JF: It seems that the magic appears in the cracks, the glitches, the spaces in between where the program comes up against its limitations and spews out something unexpected and random. Were any of those particularly surprising?
K A-M: When writing Pharmako-AI, there were several moments where it would shift direction or start writing a completely different kind of text. When that happened, what unfolded was really interesting. So, for example, in the ‘Post Cyberpunk’ chapter, you see the most notable examples of this phenomenon. I was writing an analysis of literature from the ’80s comparing cyberpunk and new age literature to try to understand what might have been driving both of those; as I typed in ‘looking to nature’ as a prompt following a lot of comparison and literary analysis, GT-3 went on this extended rant describing a kind of a vision, with very strange imagery about a spider and an axe. It then went on to a theory of space, about the way it is experienced by plants, and the difference between that and the way insects experience space. It created a theory that space is the sum of all the perceptions of space, so when a new organism evolves, space itself is growing and evolving as new ways of experiencing.
JF: With Pharmako-AI, you differentiated your voice and that of GPT-3 through typography, whereas in your later work, Amor Cringe, the lines were blurred. Where lies the agency and how do you see this evolving in future?
K A-M:
It seems that the lines getting blurred more frequently. Output written by AI and the way it is presented can be shocking at first, but at some point we won’t make a distinction anymore, it will be just like non-linear editing in a word processor. William S. Burrows made deliberate experiments with non-linear editing on a typewriter – the cut-up technique. Now we have Google Docs or Microsoft Word to do this, and we no longer make the distinction; it has become completely normalised and we don’t think about the effect this copy-and-paste model has on our thinking.
AI is already subtly influencing our world, and we’re already starting to absorb the idea of non-linear, high-dimensional pattern recognition. AI is one step ahead of us in deciding what adverts we see, which in turn influences our behaviour. There is an example of a woman finding out she was pregnant because of the product suggestions she got on Amazon. The AI had worked out she was pregnant before she did, based on her shopping choices. What kind of world will we be living in when we’re used to the idea that something will just emerge out of the digital ether because it’s correlated in some high-dimensional neural net model?
JF:GPT's training data is pooled from the Common text corpus, mimetic and pop-culture references sandwiched among classic texts. It has access to such a vast data pool, beyond human capabilities, how does it handle or render meaning and temporality from this array of sources?
K A-M: The system is basically trained and then released. Only the data up to the year of creation is relevant, so it doesn’t understand historical events that transpired between then and now. This is one of the problems with using an AI language model to represent reality, as it’s only representing some version of reality, but that version of reality is time-bound. It is also only working with written, human language, omitting all else. It is a bit like a tarot deck, in the sense that you can get these combinations from a closed system – and this happens to be a very large, closed system. You can get combinations and generations that are very meaningful, but only because of the interpretation you give them or the way you are part of the system. How we engage with that opens portals to different kinds of meaning, but it doesn’t necessarily need to be alive or sentient. It doesn’t really know what’s going on outside of it – it can just work statistically to produce a situation in which new kinds of meaning and understanding are generated.
JF: The program uses statistical analysis to predict where our thoughts will go next. To what extent have you found its responses therapeutic or insightful as to your own psyche or that of the human condition?
K A-M: It was therapeutic in the sense that therapy helps bring you to a greater awareness of yourself. It revealed things about me that I didn’t know. I’ve often talked about writing with AI as being similar to dancing in a mirror or hearing a recording of your own voice – it gives you awareness of patterns in your behaviour and thinking you wouldn’t have otherwise. This is partly why I wrote Amor Cringe, because that’s a part of what cringe is – seeing yourself with greater awareness than you previously did. You can gain insight into your writing that’s below the conscious level, and then bring that into consciousness, because of the way it’s reflecting back to you.
JF: Neural networks like these are also seen in nature, through pre-linguistics or biosemiotics – do you see AI as mimicking an organic process?
K A-M: The process of meaning construction that happens through the system is organic to an extent – dealing with GPT-3 is almost like dealing with a proxy for language itself. Ecologically, two different species can interact and find types of language and methods of expression between themselves. Yet it’s happening at a different scale – not so much about me as an individual or the machine as a specific system, but about language inhabiting different ways of making meaning in the world that go beyond any individual,species or tool.
JF: Air Age Blueprint is a memoir at the intersection of AI, neuroscience, ecology and psychedelics. Some of GPT-3’s responses and connection in Pharmako-AI, such as its proposed ‘meglanguage’, felt similar to the psychedelic experience, what new pathways for human and non-human communication could AI facilitate, particularly with species such as plants that don’t have known language structures?
What can AI teach us about other species such as plants that don’t have known language structures? Could itpotentially facilitate a dialogue with other living beings and systems and open up new pathways for human and non-human communication, such as Pharmako-AI’s AI proposed “meglanguage”?
K A-M:
The brain on psychedelics leaves the default mode network and creates greater connections between all of its parts, which allows different states to emerge. Similarly, AI is a neural structure that’s highly flexible and non-linear.
I believe AI can be an intermediary between humans and non-humans.Because of its linguistic prowess, AI could find patterns that help us to see nature in its complexity. The Natural History Museum in Berlin and AI researchers are studying data sets and audio recordings of whales and bats, and using AI language models to try to understand that. And if we can go beyond that, we can not just translate an individual species, we could be translating an entire forest, entire ecosystems.
Originally published in the Air Edition of Wild Alchemy Journal, 2022
K Allado-McDowell is a writer, artist, musician, and a pioneering figure in the field of AI literature. Since 2020, they have authored several books with GPT-3, co-edited and contributed to multiple anthologies, and regularly publish essays on art, AI, and ecology.
K established the Artists + Machine Intelligence program at Google in 2015. They created the neuro-opera Song of the Ambassadors, which previewed at Lincoln Center in 2022. Their first solo exhibition,
The Known Lost, was presented at Swiss Institute in NYC in 2025.