# Nice bit to have in the introduction - Carlson's aim at the beginning of his book. >This suggests a third model for the [[aesthetic appreciation]] of nature, [[the natural environmental]] model. [[This model]], which I develop throughout Part I of this volume, recommends two things. First, that, as in our appreciation of works of art, we must appreciate nature as what it in fact is, that is, as natural and as an environment. Second, it recommends that we must appreciate nature in light of our knowledge of what it is, that is, in light of knowledge provided by the natural sciences, especially the environmental sciences such as geology, biology, and ecology. [[The natural environmental]] model thus accommodates both the true character of nature and our normal experience and understanding of it. - we suggest that Carlson's recommendations also hold for the [[aesthetic appreciation]] of AI art (the products of generative AI). We should appreciate them for what they are, and we should appreciate them in light of our knowledge of what they are. > I think one useful way to think about neural networks is that we don't program and we don't make them. We kind of, we grow them. You know, we have these neural network architectures that we design and we have these loss objectives that we create. And the neural network architecture, it's kind of like a scaffold that the circuits grow on, and they sort of, you know, it starts off with some kind of random, you know, random things and it grows. And it's almost like the objective that we train for is this light. > > And so we create the scaffold that it grows on and we create the, you know, the light that it grows towards. But the thing that we actually create, it's this almost biological, you know, entity or organism that we're studying. And so it's very, very different from any kind of regular [[software engineering]], because at the end of the day, we end up with this artifact that can do all these amazing things. It can, you know, write essays and translate and, you know, understand images." ([[Lex Fridman]], [[Dario Amodei]]: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | [[Lex Fridman Podcast]] #452) # [[The Environmental Aesthetics of Generative AI]] ## Introduction How can we evaluate AI from an [[aesthetic point]] of view? In what follows, we suggest that function-based approaches such as [[Functional Beauty]] and [[Jane Forsey]]’s Kantian approach struggle with generative AI. We argue that adapting Carlson’s [[environmental aesthetics]] provides a [[better account]] of the appreciation of both generative AI systems and the content they produce. Carlson’s perspective—particularly his emphasis on understanding an environment in light of scientific or common-sense knowledge—can illuminate our appreciation of AI systems. By viewing these systems as environments shaped by cultural and technical forces, we can place their outputs and internal processes within a broader context that enhances aesthetic appreciation. Finally, we demonstrate how this environmental view allows for three different ways of appreciating generative AI. ## 1. Function-Based Theories Functional Beauty (Parsons and Carlson 2008) holds that an object’s aesthetic value stems from how well its form fulfills a recognisable function. Philosophers who adopt this view argue that perceptible features are best understood through their intended purpose. Jane Forsey’s Kantian approach (2013) similarly maintains that aesthetic judgment depends on how effectively an artifact’s form meets its aims. On both accounts, assessing an object aesthetically depends on understanding what it is meant to do. ## 2. Knowledge of Naturans as Naturans Carlson's natural environmental model rests on two fundamental principles, drawing an analogy from the appreciation of art. Just as appreciating art requires understanding what it is and having knowledge of its nature, so too does appreciating nature. The text states that this model recommends two things: >"First, that, as in our appreciation of works of art, we must appreciate nature as what it in fact is, that is, as natural and as an environment. Second, it recommends that we must appreciate nature in light of our knowledge of what it is, that is, in light of knowledge provided by the natural sciences, especially the environmental sciences such as geology, biology, and ecology." (p. 6) These two points form the bedrock of this model, emphasising both the _ontological_ character of nature (what it is) and the_epistemic_ element (how we know it). For the time being, let us focus on the first. ### 2.1 Nature as Naturans Works of art are intentionally created by human beings, natural objects are unintentionally created through evolutionary forces. Thought of like this, Carlson's idea that the nature of what we are appreciating feeds into the nature of our appreciation seems very plausible: why would we expect such different types of thing to be appreciated in the same way? What then is the natural environment? For Carlson, it is best understood as a dynamic, evolving whole. Consisting not simply of matter, but of the forces that act on that matter, and the systems within which that matter is. He explains that “natural objects possess what we might call an organic unity with their environments of creation,” emphasising that they are “a part of and have developed out of the elements of their environments by means of the forces at work within those environments.” To appreciate nature aesthetically, Carlson maintains that “if to aesthetically appreciate art we must have knowledge of artistic traditions and styles within those traditions, then to aesthetically appreciate nature we must have knowledge of the different environments of nature and of the systems and elements within those environments.” (p. 50) He further points out that “in the way in which the art critic and the art historian are well equipped to aesthetically appreciate art, the naturalist and the ecologist are well equipped to aesthetically appreciate nature.” (ibid.) Accordingly, recognising these processes and the interrelation of the systems and elements provides the groundwork for aesthetic appreciation. By acknowledging nature’s unity, order, and coherence in this way, we can regard it not only as an evolving physical environment but also as an aesthetically significant one. This way of conceptualising nature fits naturally with a much older idea: Spinoza's understanding of natura naturans provides an approach to further clarify this understanding of the natural environment. For Spinoza, natura naturans refers to the active, self-causing dimension of nature, which he equates with "God or Nature." In his framework, nature is not a passive repository of entities—"not a repository of pre-formed objects (natura naturata)"—but is defined by its continuous capacity to produce and maintain its own being through self-production. This aligns with Carlson's description of nature as "dynamic" and "evolving." From a Spinozist perspective, nature’s aesthetic quality does not reside in static beauty or unchanging forms but, as Carlson indicates, in the "forces and processes" and "unity, order, and harmony" that govern natura naturans’ continuous self-organising activity and the internal emergence of complexity. Consequently, the Spinozist view holds that the environmental qualities arise from the interplay of these formative forces, in close agreement with Carlson's environmental model of aesthetic evaluation. ### 2.2 Generative AI as Naturans We suggest that Carlson's recommendations also hold for the aesthetic appreciation of generative AI. We should appreciate them for what they are, and we should appreciate them in light of our knowledge of what they are. Moreover, we shall suggest in a moment that Carlson's Environmental Model (**the naturans model**) can in fact be adapted so as to provide at least the starting point for an aesthetics of generative AI. We can begin to see this by considering the following remarks from Dario Amodei, the CEO of Anthropic (my italics). > I think one useful way to think about neural networks is that we don't program and we don't make them. We kind of, *we grow them*. You know, we have these neural network architectures that we design and we have these loss objectives that we create. And the neural network architecture, it's kind of like a scaffold that the circuits grow on, and they sort of, you know, it starts off with some kind of random, you know, random things and it grows. And it's almost like the objective that we train for is this light. > > And so we create the scaffold that it grows on and we create the, you know, the light that it grows towards. But the thing that we actually create, it's this almost biological, you know, entity or organism that we're studying. And so it's very, very different from any kind of regular software engineering, because at the end of the day, *we end up with this artifact that can do all these amazing things*. It can, you know, write essays and translate and, you know, understand images." (Lex Fridman, Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452) my emphasis. The idea that LLMs are grown as opposed to programmed or made is evocative, but before considering that, it is interesting to note that if we are picky about Amodei's use of the term 'artifact,' we might wonder whether LLMs really are artifacts. On the one hand, they are intentionally brought into existence, but on the other, it is hard to understand them as having _a_ proper function. A chair is for sitting, a car is for driving, an LLM is for...? - Arguably, one ends up saying its function is something very general like producing text, producing human-like text etc, or ascribing it an extremely large number of functions. - This needs more justification? - The function of an LLM **producing human-like text** Neither is it obviously a sort of container for other artifacts, like a Swiss Army knife or a mobile phone (the function of which is arguably as an environment in which to store and run apps), as users do not install applications on an LLM, but rather they *discover* them (examples include creative writing, code generation, language tutoring, and various forms of research assistance). Many of these user innovations are quite technical, but they underscore the difficulty of assigning a single function to an LLM. >"a house is a machine for living" – To understand how LLMs might be appreciated aesthetically in a manner akin to natural environments, we can apply Carlson’s emphasis on ‘nature as naturans’ to generative systems. In the previous subsection, we saw that Carlson’s view of nature as an evolving whole, driven by interrelated processes, highlights the ways in which natural forces shape and organise matter, and that Spinoza’s notion of _natura naturans_ captures a similar idea, stressing active and continuous generation rather than static entities. ### the following needs to be written/organised better The natural environment, on Carlson’s reading, is best understood in terms of the forces that act on it and the systems these forces create. Generative AI can be interpreted similarly: its training data, alignment parameters, internal feedback loops, and algorithmic constraints act as its ‘forces.’ - Like nature, generative systems can produce varied and sometimes unpredictable ‘outputs’ that result from their ongoing processes. - The user provides prompts akin to the gardener sowing seeds, but the manner in which these prompts ‘grow’ into final outputs resembles nature’s autonomous creativity. - While the user can guide or prune the system’s generative direction, they do not directly control every detail of how the system evolves or how outputs emerge—just as a gardener does not precisely dictate how each plant develops in an ecosystem. - From this perspective, Carlson’s natural environmental model and the idea of _natura naturans_ extend to cover the generative capacity of AI, reframed as _machina naturans._ - This reflects the following points: - **Dynamic, evolving processes.** Like nature’s self-organising activity, an LLM develops its outputs through iterative training processes. - **Partial control and unpredictability.** As in gardening, users can steer but not completely determine the outcomes, echoing the incomplete control one has over a natural environment. - **Systemic interrelations.** AI models incorporate vast datasets, which shape and constrain their outputs, much like ecological forces shape natural systems. - By appreciating LLMs in this way, we adopt a stance similar to appreciating nature in Carlson’s sense. - We are encouraged to consider what generative AI _is_—namely, _machina naturans_—and to appreciate it in light of our knowledge of how it operates, how it ‘grows,’ and how our interactions iterate its generative processes. - This lays the groundwork for an environmental aesthetics of generative AI, where we foreground the system’s dynamic forces and emergent complexity rather than conceiving it as a mere artifact or tool. ### 2.3 Knowing Naturans Compost made out of experiential artifacts > we must appreciate nature as what it in fact is, that is, as natural and as an environment. Second, it recommends that we must appreciate nature in light of our knowledge of what it is, that is, in light of knowledge provided by the natural sciences, especially the environmental sciences such as geology, biology, and ecology." (p. 6) - Turning to the second half of Carlson's recommendation, if we are looking for knowledge that will allow for appreciation of ai the most obvious place to look would be in the computer sciences, all the better to understand the various algorithmic forces shaping the outputs of a generative system. - However, our contention here is that truly appreciating generative AI requires not only knowledge of the computational architecture of generative systems but also (some portion) of the multitude of experiential artifacts (images, texts) that make up a system's dataset, and, crucially, how those artifacts and their features have been encoded in the system. **Soil Analogy** - Much as the richness of soil in a garden dictates which plants flourish and how they grow, the presence or scarcity of particular artifacts in a model’s dataset shapes what styles, motifs, and creative directions it can readily generate. - By examining this “soil” of latent embeddings—understanding both the abundance of some artistic traditions and the relative absence of others—we gain a clearer sense of the system’s aesthetic potentials and constraints. **** ## 3. Approaches to Appreciating Generative AI ### maybe put something about japanese gardens and dialectics? * The core of the paradox is restated in terms of the expected difficulty of appreciating dialectical relationships. * Objects involving dialectical relationships between art and nature (like earthworks, placement pieces, topiary gardens) should be difficult and confusing to appreciate. * Japanese gardens, also involving dialectical relationships, *should* theoretically present similar appreciative difficulties. > [!NOTE] > the prompter is in dialogue with technology but also this sort of cultural mulch. ### from the old version but some good stuff in here. Building on Carlson’s environmental perspective, we can outline three overlapping ways of appreciating generative AI. First, there is theoretical appreciation. Those with a technical understanding of how generative AI works can examine the roles played by training data, algorithmic constraints, and internal feedback processes. In much the same way that scientific knowledge of geology or ecology can contribute to aesthetic appreciation of nature by revealing the structure of the natural environment, technical understanding of architectures or optimisation methods can contribute to aesthetic appreciation of AI by illuminating how an AI system’s outputs arise.  Second, there is practical appreciation, based on understanding through engaging with the generative environment. A skilled gardener may lack scientific knowledge of the environment she is in, but, through working with, and on, the environment, she develops an understanding of how such a system (its forces and objects) works. Similarly, through engaging with a generative AI, a user develops a non-technical understanding of (part of) the underlying workings of the system. In both cases, knowledge of how a generative environment’s forces function is arrived at through direct engagement, and this knowledge underlies aesthetic appreciation. Third, there is output appreciation, in which the generative environment is appreciated through the appreciation of its products. In gardens, plants reflect natural generative processes shaped by gardeners. Similarly, generative AI outputs (text, images) are shaped by data and user input. By considering these products in context, we see how both natural and computational environments generate observable outcomes that can be appreciated not only as such, but especially as expressions of the aesthetic potential of the generative AI system. ### References Carlson, A. (2005). Aesthetics and the environment: The appreciation of nature, art and architecture. Routledge. Forsey, J. (2013). The aesthetics of design. OUP USA. Parsons, G., & Carlson, A. (2008). Functional beauty. OUP Oxford. ** # The Creative Aesthetics of Generative AI ## 1. Appreciating Generative AI Systems While attention has been paid to whether and how we might appreciate the outputs of generative AI systems (AI images, music, video), this talk examines whether the systems themselves can be objects of aesthetic appreciation. Generative AI is described as a 'new type of tool', so existing accounts of design aesthetics could be extended to cover these types of objects. However, given their reliance on the concept of function, theories may not provide an aesthetics of generative AI. For Parsons and Carlson's Functional Beauty approach, understanding an object's function shapes its perceptual appearance. If an object appears well-suited to its function ("looking fit for function"), it will look aesthetically pleasing. Forsey's approach, on the other hand, assesses how an object fulfils its intended purpose, judging it as a specimen of its kind and considering how it realises its function. However, the function of generative AI systems is not obvious, and this gives us reason to doubt that either of these theories can provide a full account of the aesthetics of generative AI. First, these systems are not designed to fulfill any specific function. An LLM can be used for many things, none of which were intended by its designers while they were creating it. Second, due to their opacity and complexity, the designers of these systems are unable to explain how the form of a Large Language Model enables its various possible uses. Such functional indeterminacy means that theories like Functional Beauty and Forsey's account cannot get a foothold when attempting to explain the aesthetic appreciation of Large Language Models and other generative AI. Without a sense of function, we cannot judge how well form serves function perceptually or assess how the system realizes its purpose as a specimen of its kind. Given these difficulties, the aesthetic appreciation of generative AI may need to be approached differently. One promising perspective is "beauty in use," an aesthetic focused on the experience of interacting with objects. Beauty in use shifts the emphasis from how an object looks or fulfills its function to the qualities of the actions it enables. For instance, Nguyen suggests that we can find aesthetic pleasure in fluid, graceful interactions with tools and technologies. Similarly, generative AI systems could be appreciated for the way they facilitate creative and intellectual exploration. However, this account also faces challenges. The physical interactions we have with these systems—typing prompts or tapping a screen—are not good candidates for aesthetically pleasing actions. Moreover, a single generative AI system can have different modes of input, such as typing, voice input, or uploading an image. This variety makes it difficult for the beauty in use theory to account for their aesthetic appeal uniformly. ## 2. System Appreciation Our positive account of appreciating generative AI draws on Carlson’s environmental aesthetics, in which appreciators focus “on the order imposed on these objects by the various forces, random and otherwise, that produce them” (p. 119). On this account, and as Carlson emphasizes, appreciating an environment involves recognizing its structure and understanding the ecological and evolutionary forces that shape it. According to Carlson, understanding these forces requires a general, nonaesthetic, and nonartistic narrative that reveals the environment’s order. For example, in an old-growth forest ecosystem, the interactions between mature trees, fungi, and understory plants shows how resource exchange and mutual support drive its development. Such knowledge guides how we attend to the things around us, and correspondingly enhances aesthetic appreciation. We suggest that a similar approach can guide the aesthetic appreciation of generative AI systems. Our positive account of appreciating generative AI draws on Carlson’s environmental aesthetics, in which appreciators focus “on the order imposed on these objects by the various forces, random and otherwise, that produce them” (p. 119). On this account, and as Carlson emphasizes, appreciating an environment involves recognizing its structure and understanding the ecological and evolutionary forces that shape it. According to Carlson, understanding these forces requires a general, nonaesthetic, and nonartistic narrative that reveals the environment’s order. For example, in an old-growth forest ecosystem, the interplay among mature trees, fungi, and understory plants shows how resource exchange and mutual support drive its development. Drawing on Carlson and Ziff, this knowledge enables us to attend properly to the environment’s specific demands. Just as Venetian and Florentine paintings prompt attention to particular features grounded in their histories and natures, the forest’s ecological patterns become salient only when we know how to attend to them. Without such informed attention, appreciation collapses into mere sensory stimulation; with it, we can engage in a richer aesthetic experience informed by understanding. We suggest that Carlson's approach can be adapted to understand the appreciation of generative AI systems. This involves recognising their underlying "order," constituted by technical, mathematical, and conceptual frameworks that produce their outputs. As Carlson argues that knowledge of ecological processes informs natural environment aesthetics, knowledge of model architectures, training data, optimization processes, and emergent behaviors renders generative systems' aesthetic dimensions observable. Understanding the "forces" that produce generative content—algorithms, parameters, prompts, and datasets—allows examination of the system's patterns and stylistic signatures. One might argue such technical knowledge is inaccessible to everyday users. Unlike environmental science, communicated through documentaries and exhibits, the workings of deep neural networks may appear obscure. Most users know little about AI architectures or optimisation techniques. We suggest a response can be found in Carlson's remarks, which show that scientific knowledge and everyday practical knowledge exist on a continuum. He writes: #### "If we recognize our scientific knowledge of the natural world as only a finer-grained version of our common knowledge, then the difference between the arousal model and the natural environmental model is mainly one of emphasis." Consider the gardener who develops practical understanding without formal training. They observe which plants grow in particular conditions and recognize interactions between sunlight, moisture, and soil. This understanding parallels, though less specialized than, a botanist's knowledge. Similarly, a frequent AI user develops practical understanding through interaction. Through trial and feedback, users identify patterns in AI responses and anticipate how prompts affect outputs. This knowledge differs from a researcher's technical understanding in degree, not kind. It allows users to observe how the AI's "ecology"—its corpus, tuning procedures, and patterns—manifests in generating content. As the gardener's understanding enables aesthetic examination of nature, the prompter's practical knowledge allows observation of AI's governing "forces." The user sees a structured, evolving system responsive to input changes. Creative Aspection and AI version 2 Creative Aspection and AI The concept of acts of aspection, as introduced by Allen Carlson, provides a framework for understanding the specific perceptual and attentive activities we perform when appreciating art or nature. These activities are informed by our scientific and common-sense knowledge of what sort of object it is that we are appreciating. This paper examines how gardening and prompting generative AI function as creative acts of aspection, where engagement with a medium is actively shaped by knowledge and results in the creation of something new. 1. Acts of Aspection in Environmental Appreciation 1.1 Carlson's Concept of Acts of Aspection Carlson argues that acts of aspection involve specific perceptual activities guided by our understanding of what we are engaging with, enabling appropriate aesthetic appreciation: "In knowing the type we know what and how to appreciate. Generally speaking, a different act of aspection is performed in connection with works belonging to different schools of art, which is why the classification of style is of the essence" (Carlson, Chapter 4, p. 42). He illustrates this with examples from art, showing how different styles demand different acts of aspection. For instance, appreciating a Tintoretto involves surveying balanced masses, while appreciating a Bosch requires scanning intricate details. Paul Ziff reinforces this idea: "Different actions are involved. Do you drink brandy in the way you drink beer?" (Carlson, Chapter 4, p. 41). Carlson extends this concept to the environment, applying the idea that our knowledge informs how we appreciate nature. He provides examples of acts of aspection in different natural settings. For instance, appreciating a prairie requires surveying the open landscape, feeling the wind, and observing the subtle contours of the land—an act of aspection involving broad visual scanning and sensory engagement with openness (Carlson, Chapter 4, p. 50). In contrast, appreciating a dense forest necessitates close examination of details, attentive listening to subtle sounds, and smelling the scents of spruce and pine—an act of aspection involving detailed scrutiny and sensory focus on an intricate environment (Carlson, Chapter 4, p. 50). These acts are informed by our knowledge of the environment's characteristics, guiding us on what to focus on and how to engage. Carlson emphasises that both common-sense and scientific knowledge play crucial roles in informing these acts of aspection, highlighting specific features for appreciation and shaping our perceptual approach. Through understanding the environment's type, ecological systems, and elements, as well as common-sense knowledge such as how a prairie or forest is typically experienced, we can perform the appropriate acts of aspection, leading to a richer aesthetic experience. ### 1.2 Gardening as a Creative Act of Aspection Gardening represents a distinct form of act of aspection that involves both perception and active engagement with the environment. Unlike the acts of aspection Carlson describes, which focus on appreciation without altering the environment, gardening includes the transformation of nature through cultivation. This transformation is guided by the gardener's knowledge of plants, soil, and ecological interactions, which informs both their perceptual focus and physical actions. For example, a gardener observes the growth patterns of plants, notices subtle changes indicating nutrient deficiencies, and anticipates how plants will mature and interact spatially. These observations are not merely passive but are connected to deliberate actions that shape the garden's development. The gardener's engagement is thus a creative act of aspection, where perception and action are intertwined. Carlson acknowledges that different environments require different acts of aspection informed by knowledge (Carlson, Chapter 4, p. 50). Gardening extends this idea by demonstrating how knowledge not only guides perception but also informs creative interaction with the environment. The gardener's understanding of horticulture enables them to perform specific acts of aspection that both appreciate and modify the environment, leading to an enriched aesthetic experience that is both appreciative and productive. By actively shaping the environment, gardeners engage in a process that enhances their appreciation of nature through creation. This creative involvement allows them to experience the environment in a way that combines aesthetic appreciation with practical engagement, deepening their understanding of natural processes and their aesthetic manifestations. Gardening thus exemplifies a creative act of aspection that extends Carlson's framework by incorporating transformation alongside perception. 1. Prompting Generative AI as a Creative Act of Aspection Prompting generative AI can be conceptualized as a creative act of aspection. Just as gardening allows for a particular type of aspective act that enables the appreciation of nature, prompting generative AI enables a creative engagement with the AI system itself. Users interact with the AI system by crafting prompts informed by their knowledge of language, art styles, and the AI's capabilities. This process involves specific perceptual and conceptual activities guided by understanding, much like the acts of aspection in gardening. By engaging with the AI, users perform acts of aspection that focus their attention on how specific inputs influence outputs. This engagement is not solely about producing images or text but about appreciating the generative process of the AI system. Just as gardeners use their knowledge to direct their engagement with the garden, prompters use their understanding to shape the AI's outputs. Both practices involve an active, creative engagement where knowledge informs both perception and creation, resulting in a dynamic interaction with the medium. In prompting generative AI, users perform acts of aspection by focusing their attention on how specific inputs influence the AI's outputs. They observe the effects of different prompt phrasings, styles, and references, developing an understanding of the system's latent space. This engagement requires knowledge of artistic concepts, styles, and the AI's mechanisms, guiding the user's perceptual and conceptual focus—much like how knowledge informs acts of aspection in art appreciation. A user familiar with Impressionism, for example, might craft prompts referencing Monet to elicit certain aesthetic qualities, understanding that the AI interprets such references based on its training data. This specialized knowledge informs the user's acts of aspection, directing attention to how the AI interprets and represents artistic elements. Just as gardening involves a creative act of aspection that leads to an appreciation of nature, prompting generative AI involves a creative act of aspection that leads to an appreciation of the AI system itself. By shaping the outputs through informed prompts, users deepen their understanding and appreciation of the AI's generative capabilities. The user's engagement becomes an exploration of the medium's potential, guided by knowledge and active participation, which parallels the gardener's interaction with the garden ecosystem. The creative aspect of AI prompting distinguishes it from traditional acts of aspection in several ways. Users actively shape the outputs through their inputs, paralleling the creative acts of aspection in gardening and art creation. This contrasts with the acts of aspection Carlson describes for nature appreciation, where the observer does not alter the environment. In AI prompting, users engage in a dialogue with the system, influencing its generative process through their knowledge and choices. The act of crafting prompts becomes a method of exploring and appreciating the medium, guided by specialized knowledge similar to that used in art creation and appreciation. This creative engagement deepens the user's perception of the AI's potential for creative expression, enriching their aesthetic experience. Applying Carlson's concept of acts of aspection to AI prompting extends his framework into the digital realm. Understanding the AI system's nature and capabilities informs the user's acts of aspection, guiding their engagement. By performing creative acts of aspection through prompting, users not only produce new outputs but also appreciate the AI system as a generative medium. This parallels Carlson's emphasis on the role of knowledge in informing acts of aspection, highlighting how understanding the medium shapes the aesthetic experience. 1. Comparing Acts of Aspection in Art, Gardening, and Generative AI In discussing appreciation in art, gardening, and AI prompting, the term 'medium' better captures the essence of what is engaged with, as it encompasses the materials, tools, and processes involved. Carlson emphasizes appreciating nature "as nature," focusing on its inherent qualities and processes (Carlson, Chapter 1, p. 12). Similarly, in AI prompting and gardening, the 'medium' includes the materials and processes through which creative expression is realized. A painter performs acts of aspection through a deep engagement with their medium—paints, brushes, canvas—and the physical properties of these materials. The painter gains understanding of how materials behave, how colors blend, and how brushstrokes affect the canvas, akin to a gardener learning how plants respond to different conditions. This experiential knowledge guides the painter's creative process, allowing them to anticipate outcomes and adjust techniques accordingly. Like the gardener, the painter's acts of aspection involve both perception and creation, informed by practical, hands-on experience. However, the painter may not be appreciating an external 'medium' in the same way the gardener or AI prompter does, as the materials are tools for expressing an internal vision rather than elements of an independent environment. Gardeners engage with a living medium—the garden ecosystem—that responds autonomously to their actions. Their creative acts of aspection involve understanding and appreciating the garden's natural processes, such as growth, adaptation, and ecological interactions. The gardener learns from the garden's responses, gaining practical knowledge that guides future actions. This reciprocal relationship allows the gardener to appreciate the medium itself as an active participant in the creative process. AI prompters interact with a digital medium that generates outputs based on their inputs, displaying emergent behaviors not entirely predictable. Their acts of aspection involve understanding how the AI processes prompts and appreciating the medium's capabilities and limitations. Like the gardener, the prompter learns from the AI's responses, refining their prompts to achieve desired results. The AI serves as a medium that both influences and is influenced by the user's creative acts, enabling appreciation of its generative processes. These three domains—art creation, gardening, and AI prompting—share several fundamental similarities while maintaining distinct characteristics. All three involve creative acts of aspection that blend perception and creation, with practitioners gaining practical knowledge through direct interaction with their medium. The creative process in each domain is informed by understanding how the medium responds to their actions. However, significant differences exist in how these interactions manifest. In art creation, particularly painting, the medium remains inert and fully controlled by the artist, with knowledge being largely practical and skill-based, developed through hands-on experience. The appreciation focuses primarily on the artwork as an expression of the artist's internal vision. Gardening presents a different paradigm, where the medium exists as a living, dynamic environment with autonomous processes. Knowledge in this domain includes both practical experience and an understanding of natural processes, with appreciation extending to the garden's intrinsic qualities and its responses to cultivation. AI prompting introduces yet another dimension, where the medium functions as a responsive digital system with complex, emergent behaviors. Knowledge in this context combines practical interaction with specialized understanding of art history and AI mechanics, leading to appreciation that encompasses both the outputs and the generative processes of the AI. The types of knowledge required in each domain reflect these distinctions. Practical knowledge, developed through hands-on experience and skill refinement, proves essential in both painting and gardening, while AI prompting relies more heavily on knowledge acquired through iterative experimentation with the system. Theoretical knowledge varies in importance across domains—in painting, it may include art theory but remains secondary to practical skill; in gardening, scientific understanding of botany and ecology can enhance practice but is not always necessary; in AI prompting, specialized knowledge of art history, styles, and AI functionalities becomes crucial for effective system guidance. The engagement with medium manifests differently across these domains as well. Painters engage with physical materials to realize an internal vision, while gardeners collaborate with a living environment, balancing control and autonomy. AI prompters, in turn, interact with a digital system that interprets inputs to generate outputs. These distinctions influence how practitioners in each domain appreciate their creative process. Painters appreciate the emerging artwork as it aligns with their vision, gardeners appreciate the garden's development and natural processes, and AI prompters appreciate both the creative possibilities and the AI's generative process. Conclusion Gardening and prompting generative AI both function as creative acts of aspection involving informed, purposeful engagement with a responsive medium, paralleling but also differing from processes in art creation and appreciation. These practices extend Carlson's concept of acts of aspection by incorporating creation alongside perception and by appreciating the medium itself through creative engagement. The comparison highlights the role of different types of knowledge—practical skills for painters and gardeners, specialized theoretical knowledge for AI prompters—in shaping acts of aspection across different domains. Recognizing these distinctions emphasizes understanding the medium in performing appropriate acts of aspection and suggests that creative engagement with a responsive medium enhances aesthetic appreciation. Further exploration of these parallels could illuminate how knowledge, creativity, and medium appreciation interplay to shape aesthetic experiences across diverse contexts. ## Notes/Scrap # The Creative Aesthetics of Generative AI ## Appreciating Generative AI Systems While attention has been paid to whether and how we might appreciate the outputs of generative AI systems (AI images, music, video), this talk examines whether the systems themselves can be objects of aesthetic appreciation. Generative AI is described as a "new type of tool," so existing accounts of design aesthetics could be extended to them. However, because these theories rely on the concept of function, they may not suffice for generative AI. According to Parsons and Carlson’s Functional Beauty, if an object appears well-suited to its function, it looks aesthetically pleasing. Forsey's approach, on the other hand, assesses how an object fulfils its intended purpose, judging it as a specimen of its kind and considering how it realises its function. But the function of generative AI is not obvious. An LLM can be used for unforeseen purposes, and its designers cannot fully explain how its form enables these uses. Without a clear function, we cannot judge how form serves function or how the system realises its purpose as a specimen of its kind. ## System Appreciation Our positive account draws on Carlson’s environmental aesthetics, in which appreciators focus “on the order imposed on these objects by the various forces, random and otherwise, that produce them.” Appreciating an environment involves recognising its structure and the ecological and evolutionary forces shaping it. Such knowledge guides attention to patterns and enhances aesthetic appreciation. We suggest adapting Carlson’s approach to generative AI. Here, the “order” is formed by technical, mathematical, and conceptual frameworks—model architectures, training data, optimisation processes, and emergent behaviours. Understanding these “forces” reveals the system’s patterns and stylistic signatures. It might be objected that most users know little about AI architectures or optimisation techniques. Yet Carlson shows that scientific and everyday practical knowledge lie on a continuum, stating that "our scientific knowledge of the natural world is only a finer-grained version of our common knowledge."  Consider the gardener who develops practical understanding without formal training. They observe which plants thrive under certain conditions and notice interactions between sunlight, moisture, and soil. Though less specialised than a botanist, this knowledge is not fundamentally different. Similarly, frequent AI users gain practical understanding through interaction. By experimenting with prompts, they identify patterns in AI responses and anticipate outcomes, thereby perceiving how the AI’s “ecology”—its corpus, tuning procedures, and patterns—manifests in generating content. Just as the gardener’s know-how informs their appreciation of an ecosystem, the user’s evolving understanding of a generative AI system allows them to perceive its underlying order, thereby providing the materials for aesthetic appreciation ### References Carlson, A. (2005). Aesthetics and the environment: The appreciation of nature, art and architecture. Routledge. Consider first, the following quote from a recent interview with a research at Anthropic: "“I think one useful way to think about neural networks is that we don’t program them, and we don’t make them. We kind of, we grow them." \ They, too, present an “order imposed” by complex forces—here algorithms and data distributions rather than climate and geology—and we can understand these forces through a broad, nonaesthetic account of how such systems come into being. Just as the forest’s structure emerges over time through natural processes, a generative model’s intricate patterns develop through iterative training routines that “grow” solutions rather than follow a preset design. : In other words, the “forces” that shape a model’s internal structure are its training processes, data inputs, and architectural constraints. The “account” that makes this order intelligible need not be deeply technical, just as a forest visitor need not be an ecologist. With a basic narrative about training objectives and neural network architectures, users can perceive a generative model’s underlying order, learning to prompt, probe, and guide it as one might explore and understand the rhythms and textures of a natural environment. By seeing large language models and image generators as environments shaped by growth-like processes, we tap into Carlson’s logic: our aesthetic appreciation emerges through a practical, story-driven sense of how these complex entities are formed and sustained. Carlson's insights into how knowledge makes order appreciable can be extended beyond just artworks or environments to the appreciation of complex, dynamic systems. In other words, the underlying lesson is that certain objects—especially those not easily captured by familiar functions—can be appreciated by understanding them as systems whose complexity must be unpacked by appropriate knowledge. As Carlson notes, appreciation involves focusing on the order imposed on objects by various forces and processes. He explains that: "An individual qua appreciator selects objects of appreciation from the things around him or her and focuses on the order imposed on these objects by the various forces, random and otherwise, that produce them. Moreover, the objects are selected in part by reference to a general nonaesthetic and nonartistic story that helps make them appreciable by making this order visible and intelligible. Awareness and understanding of the entities—the order, the forces that produce it, and the account that illuminates it—and of the interplay among them dictate relevant acts of aspection and guide the appreciative response." (Carlson, p.119) While Carlson's original context involves environments and natural forces, the same logic applies to appreciating any complex system. The takeaway is that systems present patterns and orders shaped by multiple interacting elements. Whether the system is ecological, technological, or cultural, knowledge and the "stories" we tell about it (be they scientific explanations or everyday understandings) inform how we look at it and what features we consider significant. Carlson's notion that both common-sense and more refined scientific or theoretical knowledge guide our perception means that we can appreciate a system at multiple levels, depending on our familiarity and expertise. By framing generative AI systems as complex, dynamic entities—rather than simple tools or objects with a straightforward function—we open the door to a form of aesthetic appreciation that depends on knowledge-driven acts of aspection, much like how one comes to appreciate the complexity of a finely tuned machine, a musical ensemble, or a carefully designed garden. The point is recognizing that generative AI systems "make demands" on how we attend to them, and that meeting these demands requires more than looking for a stable function or person-like qualities. This perspective on system appreciation lays the groundwork for understanding how we might aesthetically appreciate generative AI. Instead of forcing AI into existing categories of art objects or personal interlocutors, we can treat them as complex systems whose outputs and behaviors invite certain ways of looking, interpreting, and understanding. This prepares us to consider more active and creative forms of engagement with the system. **