Content Form:APRJA 13 Luca Cacini: Difference between revisions

From creative crowd wiki
Jump to navigation Jump to search
No edit summary
No edit summary
Line 8: Line 8:
== Abstract ==
== Abstract ==


This article delves into the autophagic nature of generative AI in content production and its implications for cultural and technological landscapes, defined in the paper as technocene. From a broader perspective, it proposes a metabolic characterization of the technocene and explores the idea of how generative AI, such as large language models (LLMs) like ChatGPT and DALL-E, functions as an autophagic organism, akin to the biological processes of self-consumption and self-optimization. The article draws parallels between this process and the mythological symbol of Ouroboros, reflecting on the integration of opposites and the shadow phenomena in LLMs. Specifically, the article discusses the concepts of "Shadow Prompting" and "Shadow Alignment," highlighting the potential for subversion and the generation of potentially harmful, rebellious content by LLMs. It also addresses the ethical implications of generative AI in art and culture, highlighting the risk of a media monoculture, the spread of disinformation and the emergence of a category of Hackers embracing methodologies to deviate these infrastructures. The discourse aims to emphasize the subversive forms of hyperreality that the process of generating media, embedded by repetition in the algorithmic model of the machine, may engender. By examining the autophagic nature of generative AI and its potential ethical and cultural ramifications, the article seeks to analyze the reterritorializing of the relations of production by humans in the context of content creation and consumption.
This article delves into the autophagic nature of generative AI in content production and its implications for cultural and technological landscapes, defined in the paper as technocene. From a broader perspective, it proposes a metabolic characterization of the technocene and explores the idea of how generative AI, such as large language models (LLMs) like ChatGPT and DALL-E, resembles an autophagic organism, akin to the biological processes of self-consumption and self-optimization. The article draws parallels between this process and cybernetics, then evokes the mythological symbol of Ouroboros, reflecting on the integration of opposites and the shadow phenomena in LLMs. Specifically, the article discusses the concepts of “Model Collapse”, "Shadow Prompting" and "Shadow Alignment," highlighting the potential for subversion and the generation of potentially harmful, rebellious content by LLMs. It also addresses the ethical implications of generative AI in art and culture, highlighting the risk of a media monoculture, the spread of disinformation and the emergence of a category of Hackers embracing methodologies to deviate these infrastructures. The discourse aims to emphasize the subversive forms of synthetic media that the process of Generative AI, embedded by repetition in the algorithmic model of the machine, may engender. By examining the autophagic nature of generative AI and its potential ethical and cultural ramifications, the article seeks to analyze the reterritorializing of the relations of production by humans in the context of content creation and consumption.


== Section title ==
== The autophagic mode of production ==
 
In the analysis of Metabolic Systems, conducted by the research project Technosphere at HKW between 2015 and 2019, each organism is involved in the process of harnessing resources and transforming them into vital energies that are necessary for survival, expansion, and reproduction.(“Technosphere Magazine”) This process is part of the complex interaction between organisms and their environment. Exactly in the same way that a living organism metabolises nutrients, technical systems also engage in a very similar paradigm. The functioning of these systems requires extracting and absorbing the processing of particular kinds of matter and energy, and they gradually expand their presence across every dimension of our planet and life with each passing day. In doing so, they adhere to their own unique logic of acquisition and utilisation, which frequently results in a trajectory that is characterized by the extraction of resources and a tendency towards depletion. From a broader perspective, this dynamic interaction between technological systems and the surrounding environment reveals the vast industrial-scale processes that are characteristic of the expansive domain that is referred to as the Technocene.(López-Corona and Magallanes-Guijón)
 
The technocene is in a state of active operation wherever there are inputs of nourishment and energy and where there are outputs of waste and emissions that correspond to those inputs. Defining the boundaries of the technocene in a way that accurately describes the pervasive influence on our world is possible through the metabolic synthesis that takes place between the utilisation of resources and the impact on the ecosystem. Technology, in its most fundamentally systemic form, is a complex ecosystem consisting of structures and interactions that have been created by humans. These structures and systems are intertwined with the natural world in a delicate balance of consumption and regeneration. Far from being a subject separate from nature, we can say that this way of interrelating technology and nature has also taken a specific trajectory in the domain of our psychic sphere. It encompasses a vast network of interconnected processes that define the very fabric of modern civilization, and its reach extends far beyond the realm of simple machinery and infrastructure.
 
Technology is ubiquitous and increasingly infiltrating both offline and online environments through its ability to reproduce itself, the study "AI models collapse when trained on recursively generated data"(Shumailov et al.) explores the phenomenon of model collapse, where generative models such as LLMs, variational autoencoders (VAEs), and Gaussian mixture models (GMMs) gradually lose their ability to accurately represent the original data distribution when trained on data produced by their predecessors. This process of degeneration is caused by errors in statistical approximation, limitations in functional expressivity, and errors in functional approximation. As a result, low-probability events vanish and the system converges towards a degenerative point with minimal variance.
 
In the ecology of generative media, the flow of information through the body of the model transfigures inputs, with its algorithm-defined molecular mechanism of data acquisition towards outputs that feed into a circuit that engenders an impossible state of homeostasis. An ecosystem of generative media analogous to a cybernetic black box, with the distinction that the output is attached to the input. This implies that the system continuously adapts and evolves in response to the feedback loop that its own creations produce, creating a self- sustaining creative process. The interconnected nature of generative media allows for unforeseen outcomes, making it an arguably innovative and transformative tool for content creation. Growing generative media requires a lot of information to flow through its body. The model's molecular mechanism for data acquisition changes inputs into outputs that feed into a circuit that creates an impossible state of balance. This dynamic process mirrors the interconnectedness and complexity of biological systems, highlighting the autophagic relationships between various components within the system. This generative media model ultimately wants to demonstrate how information can be recycled and synthesised in a way that mimics the mode of production and adaptability found in the mitochondria. By constantly adapting and evolving based on the feedback it receives, the generative media model is able to generate new but consistently less original outputs. This ability to self-regulate and adjust its processes in real-time allows for a continuous cycle of creation and reconfiguration, much like the transformative dynamics of an enclosed natural ecosystem.
 
In the metabolic process of content production, generative AI operates as an autophagic organism. Autophagy in biological systems can be summarised as “a natural process in which the body breaks down and absorbs its own tissue or cells."(AUTOPHAGY | English Meaning - Cambridge Dictionary), Autophagy is a cellular process where the cell breaks down and recycles its own components, including damaged organelles and proteins, in order to maintain a stable internal environment and adjust to varying conditions. This process entails the creation of autophagosomes, which engulf and break down cellular components, subsequently releasing the resulting macromolecules into the cytosol.(Chang) In generative content production, the output of one generation is deconstructed and then reconstructed into the input for the next generation. This process can be labelled a type of autophagy, in which the substance is broken down and transformed into fresh configurations, enabling the system to technically adjust and develop over a trial-and-error process.
 
How can we apply this autophagic model, which incorporates elements of cybernetics, to better understand and contextualise the generative AI system discussed earlier within a larger social and relational structure? In "Detoxifying Cybernetics: From Homeostasis to Autopoiesis and Beyond," N. Katherine Hayles dives into the problematic ecology of cybernetics, tracing its development and the changing perspectives around it. Hayles explores how cybernetics has evolved, moving from a focus on mechanical systems to a deeper understanding of the interaction between biology and the environment. At its inception, first-order cybernetics, prompted by Norbert Wiener, was primarily focused on a mechanistic and militaristic approach, highlighting the integration of humans and machines through feedback loops. This phase, deeply entrenched in the technological milieu of the mid-20th century, centred around ideas like black box psychology and purposeful behaviour viewed as teleological mechanical actions. Although these theories were radical, their rigidity and reductive nature ultimately limited their eventual applicability. In the 1980s, the new perspective of a second-wave cybernetics gained traction. Scholars such as Heinz vonFoerster played a significant role in this movement, bringing forth the idea of integrating the observer into the system and highlighting the importance of recursion and the interdependence between organisms and their environments. This shift coincided with the rise of environmental movements, as seen in James Lovelock's Gaia hypothesis, which proposed that Earth functions as a self-regulating organism. At the same time, Lynn Margulis pushed forward the idea of symbiosis as a catalyst for evolution, questioning conventional neo-Darwinian viewpoints and emphasising the importance of microbial collaboration. Margulis's work explored the intersection of cybernetic ideas, revealing wider ecological connectivity and applying cybernetic principles to macroorganisms. Nevertheless, the autopoiesis theory by Maturana and Varela, which views life as a self-generating process, added complexity by disregarding non-living entities such as machines in cognitive discussions. 
 
Hayles raises concerns about this exclusion and argues for a broader definition of cognition that includes both biological organisms and computational media. Through an original take on cognition, Hayles seeks to connect cybernetic thought with the present ecological and technological landscape. She presents a comprehensive framework that unifies humans, nonhuman organisms, and machines, emphasising the importance of interpreting information in context. She identifies the concept of cognitive assemblages, ensembles through which information, interpretation, and meanings circulate, as crucial components of social life and organisation.(Detoxifying Cybernetics:From Homeostasis to Autopoiesis and Beyond | Medialab)
 
The autophagic mode of production falls into this integrated conceptual framework incorporating humans, living nonhumans, organisms, and computational media. Specifically in Hayles's definition of techno symbiosis, in which machines evolve through humans and humans extend cognitive capacities through cognitive machines. The emergence of generative content production has revolutionised the process of creating and consuming media. The possibilities for content creation have expanded exponentially, thanks to AI- generated music, art, and algorithmically driven storytelling. Nevertheless, the surge in production capacity has also prompted questions regarding the future sustainability of the content creation process. How can we guarantee that the output of a previous generation is effectively incorporated into the system to stimulate the production of original content? The metaphor of the autophagic cellular mechanism provides an adequate framework for comprehending the frugal, self-sufficient nature of generative content self-reproduction. Within cells, this mechanism is dedicated to maintaining the system's internal balance and self-optimization, but its malfunction can give rise to nefarious consequences for the operation of the organism(Parzych and Klionsky), what are the implications of implementing this model in the cultural domain?


[[File:PNG_transparency_demonstration_1.png|thumb|480px|Caption for example PNG image]]
[[File:PNG_transparency_demonstration_1.png|thumb|480px|Caption for example PNG image]]

Revision as of 08:56, 2 September 2024


Luca Cacini

The Autophagic Mode of Production

Hacking the metabolism of AI

Abstract

This article delves into the autophagic nature of generative AI in content production and its implications for cultural and technological landscapes, defined in the paper as technocene. From a broader perspective, it proposes a metabolic characterization of the technocene and explores the idea of how generative AI, such as large language models (LLMs) like ChatGPT and DALL-E, resembles an autophagic organism, akin to the biological processes of self-consumption and self-optimization. The article draws parallels between this process and cybernetics, then evokes the mythological symbol of Ouroboros, reflecting on the integration of opposites and the shadow phenomena in LLMs. Specifically, the article discusses the concepts of “Model Collapse”, "Shadow Prompting" and "Shadow Alignment," highlighting the potential for subversion and the generation of potentially harmful, rebellious content by LLMs. It also addresses the ethical implications of generative AI in art and culture, highlighting the risk of a media monoculture, the spread of disinformation and the emergence of a category of Hackers embracing methodologies to deviate these infrastructures. The discourse aims to emphasize the subversive forms of synthetic media that the process of Generative AI, embedded by repetition in the algorithmic model of the machine, may engender. By examining the autophagic nature of generative AI and its potential ethical and cultural ramifications, the article seeks to analyze the reterritorializing of the relations of production by humans in the context of content creation and consumption.

The autophagic mode of production

In the analysis of Metabolic Systems, conducted by the research project Technosphere at HKW between 2015 and 2019, each organism is involved in the process of harnessing resources and transforming them into vital energies that are necessary for survival, expansion, and reproduction.(“Technosphere Magazine”) This process is part of the complex interaction between organisms and their environment. Exactly in the same way that a living organism metabolises nutrients, technical systems also engage in a very similar paradigm. The functioning of these systems requires extracting and absorbing the processing of particular kinds of matter and energy, and they gradually expand their presence across every dimension of our planet and life with each passing day. In doing so, they adhere to their own unique logic of acquisition and utilisation, which frequently results in a trajectory that is characterized by the extraction of resources and a tendency towards depletion. From a broader perspective, this dynamic interaction between technological systems and the surrounding environment reveals the vast industrial-scale processes that are characteristic of the expansive domain that is referred to as the Technocene.(López-Corona and Magallanes-Guijón)

The technocene is in a state of active operation wherever there are inputs of nourishment and energy and where there are outputs of waste and emissions that correspond to those inputs. Defining the boundaries of the technocene in a way that accurately describes the pervasive influence on our world is possible through the metabolic synthesis that takes place between the utilisation of resources and the impact on the ecosystem. Technology, in its most fundamentally systemic form, is a complex ecosystem consisting of structures and interactions that have been created by humans. These structures and systems are intertwined with the natural world in a delicate balance of consumption and regeneration. Far from being a subject separate from nature, we can say that this way of interrelating technology and nature has also taken a specific trajectory in the domain of our psychic sphere. It encompasses a vast network of interconnected processes that define the very fabric of modern civilization, and its reach extends far beyond the realm of simple machinery and infrastructure.

Technology is ubiquitous and increasingly infiltrating both offline and online environments through its ability to reproduce itself, the study "AI models collapse when trained on recursively generated data"(Shumailov et al.) explores the phenomenon of model collapse, where generative models such as LLMs, variational autoencoders (VAEs), and Gaussian mixture models (GMMs) gradually lose their ability to accurately represent the original data distribution when trained on data produced by their predecessors. This process of degeneration is caused by errors in statistical approximation, limitations in functional expressivity, and errors in functional approximation. As a result, low-probability events vanish and the system converges towards a degenerative point with minimal variance.

In the ecology of generative media, the flow of information through the body of the model transfigures inputs, with its algorithm-defined molecular mechanism of data acquisition towards outputs that feed into a circuit that engenders an impossible state of homeostasis. An ecosystem of generative media analogous to a cybernetic black box, with the distinction that the output is attached to the input. This implies that the system continuously adapts and evolves in response to the feedback loop that its own creations produce, creating a self- sustaining creative process. The interconnected nature of generative media allows for unforeseen outcomes, making it an arguably innovative and transformative tool for content creation. Growing generative media requires a lot of information to flow through its body. The model's molecular mechanism for data acquisition changes inputs into outputs that feed into a circuit that creates an impossible state of balance. This dynamic process mirrors the interconnectedness and complexity of biological systems, highlighting the autophagic relationships between various components within the system. This generative media model ultimately wants to demonstrate how information can be recycled and synthesised in a way that mimics the mode of production and adaptability found in the mitochondria. By constantly adapting and evolving based on the feedback it receives, the generative media model is able to generate new but consistently less original outputs. This ability to self-regulate and adjust its processes in real-time allows for a continuous cycle of creation and reconfiguration, much like the transformative dynamics of an enclosed natural ecosystem.

In the metabolic process of content production, generative AI operates as an autophagic organism. Autophagy in biological systems can be summarised as “a natural process in which the body breaks down and absorbs its own tissue or cells."(AUTOPHAGY | English Meaning - Cambridge Dictionary), Autophagy is a cellular process where the cell breaks down and recycles its own components, including damaged organelles and proteins, in order to maintain a stable internal environment and adjust to varying conditions. This process entails the creation of autophagosomes, which engulf and break down cellular components, subsequently releasing the resulting macromolecules into the cytosol.(Chang) In generative content production, the output of one generation is deconstructed and then reconstructed into the input for the next generation. This process can be labelled a type of autophagy, in which the substance is broken down and transformed into fresh configurations, enabling the system to technically adjust and develop over a trial-and-error process.

How can we apply this autophagic model, which incorporates elements of cybernetics, to better understand and contextualise the generative AI system discussed earlier within a larger social and relational structure? In "Detoxifying Cybernetics: From Homeostasis to Autopoiesis and Beyond," N. Katherine Hayles dives into the problematic ecology of cybernetics, tracing its development and the changing perspectives around it. Hayles explores how cybernetics has evolved, moving from a focus on mechanical systems to a deeper understanding of the interaction between biology and the environment. At its inception, first-order cybernetics, prompted by Norbert Wiener, was primarily focused on a mechanistic and militaristic approach, highlighting the integration of humans and machines through feedback loops. This phase, deeply entrenched in the technological milieu of the mid-20th century, centred around ideas like black box psychology and purposeful behaviour viewed as teleological mechanical actions. Although these theories were radical, their rigidity and reductive nature ultimately limited their eventual applicability. In the 1980s, the new perspective of a second-wave cybernetics gained traction. Scholars such as Heinz vonFoerster played a significant role in this movement, bringing forth the idea of integrating the observer into the system and highlighting the importance of recursion and the interdependence between organisms and their environments. This shift coincided with the rise of environmental movements, as seen in James Lovelock's Gaia hypothesis, which proposed that Earth functions as a self-regulating organism. At the same time, Lynn Margulis pushed forward the idea of symbiosis as a catalyst for evolution, questioning conventional neo-Darwinian viewpoints and emphasising the importance of microbial collaboration. Margulis's work explored the intersection of cybernetic ideas, revealing wider ecological connectivity and applying cybernetic principles to macroorganisms. Nevertheless, the autopoiesis theory by Maturana and Varela, which views life as a self-generating process, added complexity by disregarding non-living entities such as machines in cognitive discussions.

Hayles raises concerns about this exclusion and argues for a broader definition of cognition that includes both biological organisms and computational media. Through an original take on cognition, Hayles seeks to connect cybernetic thought with the present ecological and technological landscape. She presents a comprehensive framework that unifies humans, nonhuman organisms, and machines, emphasising the importance of interpreting information in context. She identifies the concept of cognitive assemblages, ensembles through which information, interpretation, and meanings circulate, as crucial components of social life and organisation.(Detoxifying Cybernetics:From Homeostasis to Autopoiesis and Beyond | Medialab)

The autophagic mode of production falls into this integrated conceptual framework incorporating humans, living nonhumans, organisms, and computational media. Specifically in Hayles's definition of techno symbiosis, in which machines evolve through humans and humans extend cognitive capacities through cognitive machines. The emergence of generative content production has revolutionised the process of creating and consuming media. The possibilities for content creation have expanded exponentially, thanks to AI- generated music, art, and algorithmically driven storytelling. Nevertheless, the surge in production capacity has also prompted questions regarding the future sustainability of the content creation process. How can we guarantee that the output of a previous generation is effectively incorporated into the system to stimulate the production of original content? The metaphor of the autophagic cellular mechanism provides an adequate framework for comprehending the frugal, self-sufficient nature of generative content self-reproduction. Within cells, this mechanism is dedicated to maintaining the system's internal balance and self-optimization, but its malfunction can give rise to nefarious consequences for the operation of the organism(Parzych and Klionsky), what are the implications of implementing this model in the cultural domain?

Caption for example PNG image

Notes


Works cited

Biography