World TravellingIt’s been a while! Attending the Game Developers Conference really kept me busy, and I am soon going AFK for a week, so this is the last Playing Software newsletter until April 14th. And it’s all about world travelling! Before I write about world traveling, I need to vent. Thanks for reading Playing Software! Subscribe for free to receive new posts and support my work. The critical, skeptical discourse about Generative AI, and particularly about text-based systems, insists on reminding us that these systems are nothing but very efficient autocomplete software; that what they write is not creative but a fantasmagoria of commutations; that AI systems have no soul, no intelligence, and no human traits. All of these things are right. But hammering on that argument is wrong. Look, we get it: these are just really, really good Elizas. Why should we think about them differently? Basically, because we already are treating them differently. And I am not referring about sensationalist headlines in press about chatbots falling in love or what not. Projection is not important. I am talking about the ways in which people have appropriated these systems and figured out how to have fun with them. Generative AIs are in general not quite ethical, and releasing them in the wild in this sort of arms race to the bottom to see which tech bro outbros the other is painful, both figuratively and literally, for all those artists and writers involved. But the tools are out there, these companies have no interest in slowing down their deployment, and we have to deal with it. And we’re doing so by playing around and in doing so changing elements of our culture, perhaps forever. Look, these tools are a mess, and as I said at the beginning of this newsletter, they are based on extractivist, colonialist, imperialist logics that will only further visions of empire and labor theft. So I am not really cheering for these technologies as they are being built now, much less at the fact that they are out there without any accountability. But when I see people playing with them, I have hope. Because these generative AIs were (irresponsibly) released as playthings. And through play, we can create the culture we want, the culture we desire. Mind you, it requires an effort, and the default might be just perpetuating old and tired cultural tropes - see all the sexualization of images or deep fake revenge porn or plagiarism. But we can also play with this playthings in other ways. We can approach them not as servants, not as amplifiers of our darkest desires, but as agents of multiple possibilities. Sure, these are just stochastic parrots! But their parroting, coupled with the intention of doing something else, something new, not something that is within the space of possibility of the datasets these systems have been trained for. This implies playing with this systems, not letting them substitute us, but letting them augment us, challenge us, question us. I don’t think these systems tells us anything about human intelligence. They don’t understand anything except the character prior to the next, or the labels in their datasets. But they are really fun. And in being so, they allow us to explore this weird entanglement of what a person can do with artificial agency. These systems work when they play together with human agency, with people. They should become our playmates, companions in finding new ways of expression and thinking and creating. Incredible new phenomena are happening all the time in this space, and just because these tools are simpler than what they seem, just because some hyperbolic ghouls are misleading us because they are hustling for their next VC round and fat payout, it does not mean that generative AI is less of an exciting, cultural phenomenon. It is not the Internet, or the WWW, or the iPhone. But it sure is something that is changing the world in real time. The question is, how should we approach them? If the default is to create exploitative entanglements with these systems, how can we think otherwise? By the end of the day, I think Playing Sofware is an ethics book. Maybe not a proper philosophical ethics book, but one in which the actual point is not, as I write, to make sense of the role of playing in computational culture, but to make sense of what playing can tell us, and contribute with, to living ethically with software. I think that in order to live ethically with software, we need to play with it. But not playing in the sense of playing games. Playing in the sense of a mode of relating to the rule-based, world-creating capacities of software. In a previous entry I wrote about how ChatGPT or Stable Difussion have a model of the world they act upon. To put it briefly, they have a world. In that world, they are agents, and so when we interact with software we travel to that world, we make it our shared world, so software can act and we can interact with it. Games offer a good illustration of this: if we want to play soccer, we first need to draw some lines, decide what the goals are, and find a ball. We are creating a world. In that world, our agency relates to that of the ball, and of all the other players on the pitch (and the length of the grass, etc). We approach software through play because we see in it the same type of world-creation and reformulation of agency that we see in other playthings, from games to toys. And of course, systems like ChatGPT are interfaced through natural language, our prime world-creation and representation instrument. Prompting is word play being broadcast to the world of software. The question is then: how should we play with software? In a future post I will go in a bit of details regarding the problems that drawing on “classic” play theory can create when think about an ethos of playing. Before that, let me introduce the notion of world traveling. Argentinian philosopher María Lugones wrote an article called Playfulness, “World”-Travelling, and Loving Perception in which she proposes a concept of playfulness that is extremely useful for thinking about the ethos of playing software. Lugones writes:
Once again, games can help us think productively about this: when we play a game, we often travel to that other world, we often become a different person. In The Ethics of Computer Games I wrote about this capacity for creating another self, another subject, as being the key to understand why we can play games that appear to be morally disgusting, but that are somehow reasonable for us to engage with because we are playing. Playing games is world traveling, and games are the priviliged world-making technology in our cuture of play. But we can world travel beyond games. Interacting with software is a form of world travelling. Software contains a world, and when we relate to software agency, when we use it or when it uses us (as data points, as input providers), we are travelling to its world. There are ways of framing this world traveling thinking about human experience: as work, for example, when we interact with “productivity” software. Or as part of our gregarious efforts, another element in our social relations. Or we can play. In fact, we always play with software. Because play is world travelling, is a mode of engaging with other worlds, relating to them, constructing who we are in those worlds. But we should be careful! We often understand playing, especially as adults, through the lens of competition. And that lens is inherently problematic:
Because software is so good at acting with numbers, our general default is to think about playing with software in the context of competitive play. However, competition is troublesome because it fosters a form of relationality that encourages control and dominance. Lugones proposes an alternative, an understanding of play and playfulness that is rooted in curiosity and love towards the world we travel to and the others we relate to there. And this form of loving world traveling has quite specific characteristics:
The irony of generative AI systems is that they have been rushed to the world absolutely open to this mode of playfulness. They care little about rules, and rules are not sacred; they can be humorous and are absolutely open for expression. They are not self-important, and they just generate text for us, the entangled players, to act with. The responsibility, the ethos of that relation, is with us: how we construct our relation, how we prompt the machine, what we ask it to perform and how to act and how to behave. These systems are fun, but they are also demanding. We can approach them as instruments, as tools for tasks they are not good at, and then they break and challenge our assumptions and pollute our world. But we can approach their worlds playfully. In doing so, we are responsible for whatever playing results, but we are also benefitting from the pleasures of understanding the agency of software. There’s a future newsletter about the ethics, or the ethos of world-travelling. So let me end just travelling. I think the reason we can see some form of promise in generative AI is not related to how they can substitute jobs, or how they will save us time, or how they will teach us something about the brain (they won’t). Generative AI is a fun way of travelling to the world of an alien agency that can write, can draw, can hallucinate, can follow orders and can break orders. It’s almost as if it wants us to just be playful. And yet, I am writing these final paragraphs right after the announcement that GPT has now plugins. Great! Except, not really. These plugins are basically approved fronts for corporate conversational agents. They are not forms of world-traveling to GPT’s world, but a form of engaging with the corporate world of its masters. And that is less fun. Capitalism will kill playing with generative AI. At least, with these large playthings they unwittingly released to the world. All in the name of value and growth and profit. But the actual value these systems relies on their brokenness, on their noise and distortion and hallucinations; on the prompt awaiting our instructions, on the multiple possibilities that are open in CHatGPT or DALL-E just because the rules are vague, the purposes are not clear, and we can just try to have fun. Every plugin, every application of these systems to “improve our lives” are closing down avenues towards a form of world travelling that is dangerous and exhilarating and meaningful and valuable, or, in other words, playful. Thanks for reading Playing Software! Subscribe for free to receive new posts and support my work. |
World Travelling
World TravellingIt’s been a while! Attending the Game Developers Conference really kept me busy, and I am soon going AFK for a week, so this is the last Playing Software newsletter until April 14th. And it’s all about world travelling! Before I write about world traveling, I need to vent. Thanks for reading Playing Software! Subscribe for free to receive new posts and support my work. The critical, skeptical discourse about Generative AI, and particularly about text-based systems, insists on reminding us that these systems are nothing but very efficient autocomplete software; that what they write is not creative but a fantasmagoria of commutations; that AI systems have no soul, no intelligence, and no human traits. All of these things are right. But hammering on that argument is wrong. Look, we get it: these are just really, really good Elizas. Why should we think about them differently? Basically, because we already are treating them differently. And I am not referring about sensationalist headlines in press about chatbots falling in love or what not. Projection is not important. I am talking about the ways in which people have appropriated these systems and figured out how to have fun with them. Generative AIs are in general not quite ethical, and releasing them in the wild in this sort of arms race to the bottom to see which tech bro outbros the other is painful, both figuratively and literally, for all those artists and writers involved. But the tools are out there, these companies have no interest in slowing down their deployment, and we have to deal with it. And we’re doing so by playing around and in doing so changing elements of our culture, perhaps forever. Look, these tools are a mess, and as I said at the beginning of this newsletter, they are based on extractivist, colonialist, imperialist logics that will only further visions of empire and labor theft. So I am not really cheering for these technologies as they are being built now, much less at the fact that they are out there without any accountability. But when I see people playing with them, I have hope. Because these generative AIs were (irresponsibly) released as playthings. And through play, we can create the culture we want, the culture we desire. Mind you, it requires an effort, and the default might be just perpetuating old and tired cultural tropes - see all the sexualization of images or deep fake revenge porn or plagiarism. But we can also play with this playthings in other ways. We can approach them not as servants, not as amplifiers of our darkest desires, but as agents of multiple possibilities. Sure, these are just stochastic parrots! But their parroting, coupled with the intention of doing something else, something new, not something that is within the space of possibility of the datasets these systems have been trained for. This implies playing with this systems, not letting them substitute us, but letting them augment us, challenge us, question us. I don’t think these systems tells us anything about human intelligence. They don’t understand anything except the character prior to the next, or the labels in their datasets. But they are really fun. And in being so, they allow us to explore this weird entanglement of what a person can do with artificial agency. These systems work when they play together with human agency, with people. They should become our playmates, companions in finding new ways of expression and thinking and creating. Incredible new phenomena are happening all the time in this space, and just because these tools are simpler than what they seem, just because some hyperbolic ghouls are misleading us because they are hustling for their next VC round and fat payout, it does not mean that generative AI is less of an exciting, cultural phenomenon. It is not the Internet, or the WWW, or the iPhone. But it sure is something that is changing the world in real time. The question is, how should we approach them? If the default is to create exploitative entanglements with these systems, how can we think otherwise? By the end of the day, I think Playing Sofware is an ethics book. Maybe not a proper philosophical ethics book, but one in which the actual point is not, as I write, to make sense of the role of playing in computational culture, but to make sense of what playing can tell us, and contribute with, to living ethically with software. I think that in order to live ethically with software, we need to play with it. But not playing in the sense of playing games. Playing in the sense of a mode of relating to the rule-based, world-creating capacities of software. In a previous entry I wrote about how ChatGPT or Stable Difussion have a model of the world they act upon. To put it briefly, they have a world. In that world, they are agents, and so when we interact with software we travel to that world, we make it our shared world, so software can act and we can interact with it. Games offer a good illustration of this: if we want to play soccer, we first need to draw some lines, decide what the goals are, and find a ball. We are creating a world. In that world, our agency relates to that of the ball, and of all the other players on the pitch (and the length of the grass, etc). We approach software through play because we see in it the same type of world-creation and reformulation of agency that we see in other playthings, from games to toys. And of course, systems like ChatGPT are interfaced through natural language, our prime world-creation and representation instrument. Prompting is word play being broadcast to the world of software. The question is then: how should we play with software? In a future post I will go in a bit of details regarding the problems that drawing on “classic” play theory can create when think about an ethos of playing. Before that, let me introduce the notion of world traveling. Argentinian philosopher María Lugones wrote an article called Playfulness, “World”-Travelling, and Loving Perception in which she proposes a concept of playfulness that is extremely useful for thinking about the ethos of playing software. Lugones writes:
Once again, games can help us think productively about this: when we play a game, we often travel to that other world, we often become a different person. In The Ethics of Computer Games I wrote about this capacity for creating another self, another subject, as being the key to understand why we can play games that appear to be morally disgusting, but that are somehow reasonable for us to engage with because we are playing. Playing games is world traveling, and games are the priviliged world-making technology in our cuture of play. But we can world travel beyond games. Interacting with software is a form of world travelling. Software contains a world, and when we relate to software agency, when we use it or when it uses us (as data points, as input providers), we are travelling to its world. There are ways of framing this world traveling thinking about human experience: as work, for example, when we interact with “productivity” software. Or as part of our gregarious efforts, another element in our social relations. Or we can play. In fact, we always play with software. Because play is world travelling, is a mode of engaging with other worlds, relating to them, constructing who we are in those worlds. But we should be careful! We often understand playing, especially as adults, through the lens of competition. And that lens is inherently problematic:
Because software is so good at acting with numbers, our general default is to think about playing with software in the context of competitive play. However, competition is troublesome because it fosters a form of relationality that encourages control and dominance. Lugones proposes an alternative, an understanding of play and playfulness that is rooted in curiosity and love towards the world we travel to and the others we relate to there. And this form of loving world traveling has quite specific characteristics:
The irony of generative AI systems is that they have been rushed to the world absolutely open to this mode of playfulness. They care little about rules, and rules are not sacred; they can be humorous and are absolutely open for expression. They are not self-important, and they just generate text for us, the entangled players, to act with. The responsibility, the ethos of that relation, is with us: how we construct our relation, how we prompt the machine, what we ask it to perform and how to act and how to behave. These systems are fun, but they are also demanding. We can approach them as instruments, as tools for tasks they are not good at, and then they break and challenge our assumptions and pollute our world. But we can approach their worlds playfully. In doing so, we are responsible for whatever playing results, but we are also benefitting from the pleasures of understanding the agency of software. There’s a future newsletter about the ethics, or the ethos of world-travelling. So let me end just travelling. I think the reason we can see some form of promise in generative AI is not related to how they can substitute jobs, or how they will save us time, or how they will teach us something about the brain (they won’t). Generative AI is a fun way of travelling to the world of an alien agency that can write, can draw, can hallucinate, can follow orders and can break orders. It’s almost as if it wants us to just be playful. And yet, I am writing these final paragraphs right after the announcement that GPT has now plugins. Great! Except, not really. These plugins are basically approved fronts for corporate conversational agents. They are not forms of world-traveling to GPT’s world, but a form of engaging with the corporate world of its masters. And that is less fun. Capitalism will kill playing with generative AI. At least, with these large playthings they unwittingly released to the world. All in the name of value and growth and profit. But the actual value these systems relies on their brokenness, on their noise and distortion and hallucinations; on the prompt awaiting our instructions, on the multiple possibilities that are open in CHatGPT or DALL-E just because the rules are vague, the purposes are not clear, and we can just try to have fun. Every plugin, every application of these systems to “improve our lives” are closing down avenues towards a form of world travelling that is dangerous and exhilarating and meaningful and valuable, or, in other words, playful. Thanks for reading Playing Software! Subscribe for free to receive new posts and support my work. |