English

Computation has become a medium for art, and in becoming one it has raised old questions about creativity, beauty, and authorship in new and sharp forms. When an artist writes a program that generates an image, when a composer specifies an algorithm that produces music, when a designer uses computational tools to explore forms no hand could draw, computation is functioning as an artistic medium — not merely a tool for producing predetermined results but a generative system whose outputs can surprise even their creator. And when a generative AI model produces an image or a piece of music from a text prompt, the questions become acute: is this art, who is its author, can a machine be creative, and what happens to human artists when machines can produce work that resembles theirs? The relationship between computer science and the arts is where these questions are being confronted, not abstractly but in practice, by artists and technologists making and arguing about actual work.

The relationship differs from the others in this chapter in that it is as much about practice and culture as about theory. There is a body of theory — the attempts to formalize aesthetics, the computational study of creativity, the philosophy of art confronting computation — and it matters. But the relationship lives substantially in the work: the generative art that has been made for decades, the computational design that shapes the built and digital world, the music composed algorithmically, and now the explosion of AI-generated media that has made the questions urgent for everyone, not just for the artists and theorists who were exploring them before. The boundary between computer science and art is a place of making as much as of thinking, and the thinking is often done through the making.

This section examines the relationship through three lenses: computation as an artistic medium (the generative and computational arts), the formalization of aesthetics and creativity (whether beauty and creativity can be computed), and the contemporary confrontation between generative AI and the arts. It connects to computer graphics (§6.1) and human-computer interaction (§6.2), which provide the technical foundations, and to generative AI (§5.3, §5.5), but its concern is the relationship to art and design as disciplines — what computation means for creativity, aesthetics, and authorship.

Background: Computer graphics (§6.1), HCI and design (§6.2), and generative AI (§5.3, §5.5) provide the computational side; art, design, music, and aesthetics provide the creative and critical side.

The Relationship: Computation as Creative Medium and Creative Question

Computation as an Artistic Medium

The first face of the relationship is computation as a medium for making art — a medium with properties no prior medium had, which has produced forms of art that could not exist without it. The defining property is generativity: a computational artwork can be a system that produces outputs, rather than a fixed artifact, and the system can produce results its creator did not specify in detail and could not fully predict. This shifts the artist’s role from making the artwork to making the system that makes the artwork, which is a genuinely new artistic relationship.

Generative art — art produced by a system, often with autonomy or randomness, set in motion by the artist — predates the computer (there are precedents in the use of chance and rule-based procedures by earlier artists) but found its natural medium in computation. The artist specifies a process — rules, algorithms, parameters, sources of randomness — and the process generates the work, which may be a single output, an infinite stream of variations, or an interactive system that responds to viewers. The artist’s creative act is the design of the generative system, and the aesthetic interest lies partly in the outputs and partly in the relationship between the simple specification and the complex results — the way elaborate, surprising, beautiful results can emerge from compact rules. This is an aesthetic that computation made available: the beauty of emergence, of complex results from simple generative processes, which is also a theme in the science of complex systems and which computation lets artists explore directly.

The range of computational art is wide and established. Algorithmic and generative visual art (from the early plotter drawings of the 1960s through the contemporary generative art scene, including the blockchain-based generative art that found a market) explores the visual possibilities of generative systems. Computer music and algorithmic composition (from the early experiments to contemporary live coding, where performers write code that generates music in real time before an audience) bring generativity to sound. Creative coding (the use of programming as an expressive medium, supported by tools like Processing and openFrameworks designed for artists) has made computation accessible as an artistic medium to a broad community. Computational design (the use of computation to generate and explore designs, from generative typography to parametric architecture to the generative design tools that explore vast spaces of engineering solutions) brings the generative approach to design. Across these, computation functions as a medium with its own properties — generativity, interactivity, the capacity for emergence, the ability to explore vast possibility spaces — that artists and designers use expressively.

The Formalization of Aesthetics and Creativity

The second face is the attempt to formalize aesthetics and creativity — to ask whether beauty and creativity, long regarded as quintessentially human and resistant to formalization, can be captured computationally. This is where the relationship engages the deepest questions, because aesthetics and creativity touch on what is distinctively human, and the computational engagement with them is a probe of whether that distinctiveness is as absolute as it seems.

The formalization of aesthetics has a long history that computation extended. There have always been attempts to find principles of beauty — proportion, symmetry, the golden ratio, the mathematical structure in music (harmony as ratio, since Pythagoras) — and these suggest that aesthetic response has some formal structure that might be captured. Computation allows this to be pursued: can the features that make an image or a piece of music aesthetically pleasing be identified and measured? Birkhoff’s “aesthetic measure” (1933) was an early formal attempt (beauty as a ratio of order to complexity); contemporary work uses machine learning to predict aesthetic judgments, with some success, suggesting that aesthetic response is at least partly capturable by computational models, even if what they capture is not the whole of it. The partial success is itself interesting: it suggests that aesthetics is neither purely formalizable (the models do not capture everything) nor entirely beyond formalization (they capture something real), which is a finding about the nature of aesthetic response that the computational engagement made possible.

Computational creativity — the study of whether and how machines can be creative — is the more provocative engagement. It asks what creativity is (a question that turns out to be hard to answer precisely), whether the products of computational systems can be genuinely creative or only apparently so, and what creativity in a machine would require. Margaret Boden’s analysis distinguished kinds of creativity — combinational (novel combinations of familiar ideas), exploratory (exploring the possibilities within a style or conceptual space), and transformational (changing the space itself, the deepest and rarest kind) — and this framework lets the question be asked precisely: computational systems are clearly capable of combinational and exploratory creativity, and the question of whether they can achieve transformational creativity (genuinely changing the conceptual space rather than exploring within it) is the deep one. The field has produced systems that generate art, music, and ideas that observers judge creative, which raises the question of whether the creativity is real or attributed — whether creativity is a property of the process or a judgment by an observer, a question that bears on what creativity fundamentally is.

Generative AI and the Arts: The Contemporary Confrontation

The third face is the confrontation, now urgent, between generative AI and the arts. The arrival of AI systems that generate images, music, text, and video of high quality from simple prompts (§5.3, §5.5) has transformed the relationship from a specialist concern into one that confronts every artist and that raises the old questions — about creativity, authorship, and the value of human art — in their sharpest form.

The questions are several and genuinely hard. The creativity question returns with force: when a diffusion model generates a striking image, is it creative, and if not, why not, given that the output may be indistinguishable from what we would call creative if a human made it? This presses on whether creativity is about the process (and the model’s process is not human creativity) or the product (and the product may be indistinguishable), and the difficulty of answering reveals how unclear our concept of creativity is. The authorship question is practical and contested: who is the author of an AI-generated work — the person who wrote the prompt, the developers of the model, the artists whose work trained it, or no one? The legal dimension (§9.6) intersects here, but the deeper question is what authorship means when the work emerges from a model trained on millions of human works. The training question is acute and bitter: generative models are trained on the work of human artists, often without consent or compensation, and then produce work that competes with those artists, which has produced genuine grievance and the charge that the technology is built on appropriation.

The deepest question is what generative AI means for human artistic practice and value. One view is that AI is a new tool, an extension of the long history of technology in art (photography, synthesizers, digital tools each provoked similar anxieties and each became part of artistic practice), and that artists will incorporate it as they incorporated prior technologies, using it expressively while human creativity remains central. The other view is that this technology is different — that it does not merely assist human creativity but substitutes for it, automating the production of the cultural artifacts that human artists made, devaluing human creative work and threatening the livelihoods and the meaning of artistic practice. The truth is genuinely uncertain and probably varies by domain and by how the technology develops, and the confrontation is being worked out now, in the practice of artists adopting or resisting the tools, in the markets for creative work, in the law, and in the culture’s evolving sense of what it values about human art. This is the relationship at its most consequential and most contested, and its resolution will shape both the arts and the role of computation in culture.

What This Perspective Changes

Seeing the relationship between computer science and the arts changes how a practitioner understands computation, creativity, and the cultural stakes of the technology.

The first change is recognition of computation as an expressive medium. The practitioner who sees that computation is not only a tool for building functional systems but a medium for art — with its own expressive properties of generativity, emergence, and interactivity — understands a dimension of their field that the purely functional view misses. Computation can be used to make things of beauty and meaning, and the generative and computational arts are a serious domain where this is done. For a computationally trained person, this opens a creative use of their skills that the engineering framing does not suggest.

The second change is a more examined understanding of creativity and aesthetics. The practitioner who engages with the formalization of aesthetics and the study of computational creativity confronts questions about what creativity and beauty are — questions that the partial success of computational models makes concrete rather than abstract. They see that aesthetics is partly but not wholly formalizable, that creativity is hard to define precisely and that the difficulty is revealing, and that these quintessentially human capacities are probed, if not captured, by the computational engagement. This is a deeper understanding of the human capacities than either uncritical mystification or reductive dismissal provides.

The third change is the ability to think clearly about generative AI and the arts. The practitioner who understands the confrontation can engage the urgent questions — about whether AI is creative, who authors AI-generated work, what the training of models on human art means, and what the technology means for human artistic practice — with awareness of what makes each question hard and what is genuinely at stake. This protects against both the dismissal (it is just a tool, nothing important is happening) and the catastrophizing (human art is over), and it engages the genuine, unresolved, consequential questions the technology raises.

The fourth change is appreciation of the cultural stakes. The practitioner who sees that the relationship between computation and the arts bears on what a culture values about human creativity, on the livelihoods of artists, and on the character of the cultural artifacts a society produces, understands that the technology they build has cultural consequences beyond its functional effects. The generative AI that produces images and music is not only a technical achievement but a cultural force, and the practitioner who appreciates this engages the technology’s development with awareness of stakes that the purely technical view does not register.

Resources

Foundational and Bridging Texts

Margaret Boden’s The Creative Mind: Myths and Mechanisms (2nd ed., Routledge, 2004) is the foundational work on computational creativity, providing the analysis of kinds of creativity (combinational, exploratory, transformational) that lets the question of machine creativity be asked precisely. It is the essential text for thinking rigorously about whether and how machines can be creative, by the philosopher who did most to make the question tractable.

For computation as an artistic medium, Processing: A Programming Handbook for Visual Designers and Artists (Reas and Fry, the creators of Processing) and the broader creative-coding literature provide the practical entry to making computational art, while books on generative art (Matt Pearson’s Generative Art, and the writing of practitioners) cover the aesthetic and practical dimensions. The history is captured in works on the history of computer art, documenting the decades of practice from the early plotter art onward.

For the formalization of aesthetics, the literature runs from Birkhoff’s Aesthetic Measure (1933) through contemporary computational aesthetics; the field is scattered across art theory, psychology, and computer science, and the surveys of computational aesthetics provide the entry. For the philosophy of art confronting computation, the aesthetics literature (connecting to §9.9) provides the conceptual frameworks.

Resource Role Type
Boden, The Creative Mind: Myths and Mechanisms (2nd ed.) Computational creativity; the essential analysis Depth
Reas & Fry, Processing + creative coding literature Computation as artistic medium; practical Practice
Pearson, Generative Art Generative art, aesthetic and practical Entry
Birkhoff, Aesthetic Measure / computational aesthetics surveys Formalization of aesthetics Depth

Practice and Current Sources

Computation as an artistic medium is learned by making. The creative-coding tools — Processing and p5.js (the JavaScript version, browser-based and accessible), openFrameworks, TouchDesigner for interactive and visual work, SuperCollider and the live-coding environments (TidalCycles, Sonic Pi) for music — are the means of making computational art, and the communities around them (the creative coding community, the generative art scene, the algorithmic music community) are where the practice lives. Making generative work — even simple pieces — is the way to understand computation as a medium from the inside.

For the contemporary confrontation between generative AI and the arts, the discourse is current and distributed: the writing of artists confronting AI (both those adopting it and those resisting it), the critical and theoretical writing on AI art, the legal developments (§9.6), and the ongoing public argument. This is a domain where staying current means following the discourse rather than reading settled treatments, because the questions are being worked out now. The generative AI tools themselves (the image, music, and text generators of §5.3 and §5.5) are part of the relevant experience — using them, and reflecting on what using them reveals about creativity and authorship, is part of engaging the questions.

For the deeper philosophical questions about art, creativity, and computation, the philosophy of art and aesthetics (connecting to §9.9) provides the frameworks, and the specific literature on the philosophy of AI art is developing.

Resource Role Type
Processing / p5.js / openFrameworks (free) Creative coding tools; making computational art Practice
p5.js reference and examples (free) Modern browser-based creative coding entry Practice
The Coding Train (free; optional paid support) Project-based creative coding videos Auxiliary
OpenProcessing (free; paid Plus options) Creative coding community and sketch archive Practice
SuperCollider / TidalCycles / Sonic Pi (free) Algorithmic and live-coding music Practice
Generative AI tools (§5.3, §5.5) The contemporary medium, experienced directly Practice
Contemporary AI-art discourse (ongoing) The live confrontation Reference

Traps

Trap Why it misleads Better response
Dismissing computational art as not real art The view that art made with computation — generative art, algorithmic music, AI-assisted work — is not real art, because real art requires a human hand or because the computer does the work, misunderstands both the art and the medium. Computational art involves real artistic choices (the design of the generative system, the curation of outputs, the expressive use of the medium’s properties), and the dismissal usually reflects unfamiliarity with the medium rather than a defensible aesthetic position. Every new medium has faced this dismissal (photography, electronic music) and been wrong. Engage computational art as a medium with its own properties and its own artistic practice. The artist working in generative systems makes real creative choices, and the medium’s properties (generativity, emergence, interactivity) are expressive resources. Judge the work as art, on its aesthetic merits, rather than dismissing the medium. The history of art is a history of new media being dismissed and then incorporated.
Assuming creativity and aesthetics are either fully formalizable or wholly beyond formalization Two opposite errors: the reductive view that beauty and creativity are just computable functions waiting to be found, and the mystical view that they are wholly beyond any formal understanding. The evidence from computational aesthetics and creativity points between: these capacities are partly formalizable (models capture something real) and not wholly so (they do not capture everything), and the partial success is itself the interesting finding. Hold the middle that the evidence supports: aesthetics and creativity have formal structure that computation can partly capture, and a residue that current formalization does not reach. The partial success of computational models of aesthetic judgment and the real-but-limited creativity of computational systems are findings about the nature of these capacities — neither fully mechanical nor fully mysterious — and the interesting questions are about exactly what is and is not captured.
Settling the question of whether AI is creative The question of whether generative AI is creative is treated by some as obviously yes (it produces creative-seeming work) and by others as obviously no (it is just statistics, it cannot really create), and both treat as settled a question that is genuinely hard and that reveals how unclear our concept of creativity is. The difficulty of the question is the point: it exposes that we do not have a clear account of what creativity is. Treat the question as the genuinely hard and revealing one it is. Boden’s distinctions (combinational, exploratory, transformational creativity) let it be asked precisely; the question of whether AI achieves transformational creativity is open and deep. The difficulty of answering whether AI is creative is itself informative about how unclear the concept of creativity is, and engaging that difficulty is more valuable than a confident answer in either direction.
Ignoring the stakes for human artists The questions about AI and art can be discussed abstractly — about creativity and authorship — in a way that ignores the concrete stakes for human artists, whose work trains the models, whose livelihoods the technology may threaten, and whose grievance about uncompensated use of their work is genuine. Treating the relationship as purely an intellectual puzzle ignores that real people are affected, sometimes harmed, by how it develops. Hold the intellectual questions together with the human stakes. The training of models on artists’ work without consent or compensation is a genuine grievance; the threat to artists’ livelihoods is real; the cultural stakes of automating creative production are significant. Engaging the relationship responsibly means attending to these concrete consequences for the people affected, not only to the abstract questions about creativity and authorship.

中文

计算已经成为一种艺术媒介；而在成为艺术媒介的过程中，它以新的、尖锐的形式重新提出了关于创造力、美和作者身份的古老问题。当艺术家编写一个生成图像的程序，当作曲家指定一个生成音乐的算法，当设计师使用计算工具探索人手无法画出的形式时，计算就正在作为艺术媒介发挥作用——它不仅仅是生产预定结果的工具，而是一个生成系统，其输出甚至能让创作者本人感到意外。而当生成式 AI 模型根据文本提示生成图像或音乐时，问题就变得更加尖锐：这是艺术吗？谁是作者？机器能否具有创造力？当机器能够生产类似人类艺术家的作品时，人类艺术家会发生什么？计算机科学与艺术的关系，正是这些问题被面对的地方；不是抽象地面对，而是在实践中，由正在创作和争论真实作品的艺术家与技术工作者共同面对。

这种关系不同于本章其他关系之处在于，它既关乎实践和文化，也关乎理论。确实存在一套理论：对美学形式化的尝试，对创造力的计算研究，以及艺术哲学面对计算时提出的问题；这些理论很重要。但这种关系在很大程度上存在于作品之中：几十年来已经被创作出来的生成艺术，塑造建筑世界与数字世界的计算设计，用算法作曲的音乐，以及如今 AI 生成媒体的爆发——它让这些问题不再只是早先探索它们的艺术家和理论家的关切，而是变成所有人都必须面对的紧迫问题。计算机科学与艺术的边界，既是制作之地，也是思考之地；而这种思考常常正是通过制作完成的。

本节将通过三个视角考察这种关系：作为艺术媒介的计算，即生成艺术与计算艺术；美学和创造力的形式化，即美与创造力能否被计算；以及生成式 AI 与艺术之间的当代对峙。本节连接到计算机图形学（§6.1）和人机交互（§6.2），它们提供技术基础；也连接到生成式 AI（§5.3，§5.5）。但本节真正关心的是计算与艺术、设计这些学科之间的关系——计算对于创造力、美学和作者身份意味着什么。

背景知识：计算机图形学（§6.1）、HCI 与设计（§6.2），以及生成式 AI（§5.3，§5.5）提供计算一侧；艺术、设计、音乐和美学提供创作与批评一侧。

这种关系：计算作为创造媒介与创造问题

计算作为艺术媒介

这种关系的第一种面向，是计算作为制作艺术的媒介——一种具有以往任何媒介都不具备之性质的媒介，并且已经产生出没有它就无法存在的艺术形式。它的定义性特征是生成性：一件计算艺术作品可以是一个产生输出的系统，而不是一个固定作品；这个系统可以产生创作者并未详细指定、也无法完全预测的结果。这改变了艺术家的角色：艺术家不再只是制作艺术品，而是制作那个会制作艺术品的系统。这是一种真正新的艺术关系。

生成艺术——由系统产生的艺术，通常带有一定自主性或随机性，由艺术家启动——早于计算机出现。早期艺术家使用偶然性和基于规则的程序时，就已经有其先例。但它在计算中找到了天然媒介。艺术家指定一个过程：规则、算法、参数、随机性来源；然后这个过程生成作品。作品可以是单个输出，也可以是无限变化的流，也可以是回应观众的交互系统。艺术家的创造行为在于设计这个生成系统；其美学兴趣则部分存在于输出中，部分存在于简单规则与复杂结果之间的关系中——也就是复杂、意外、美丽的结果如何能从紧凑规则中涌现出来。这是一种由计算开启的美学：涌现之美，简单生成过程产生复杂结果之美。这同样也是复杂系统科学中的主题，而计算使艺术家能够直接探索它。

计算艺术的范围很广，而且已经相当成熟。算法视觉艺术与生成视觉艺术——从 1960 年代早期的绘图仪作品，到当代生成艺术场景，包括通过区块链生成艺术获得市场的作品——探索的是生成系统的视觉可能性。计算机音乐和算法作曲——从早期实验到当代 live coding，也就是表演者在观众面前实时编写生成音乐的代码——把生成性带入声音。创意编程，也就是把编程作为表达媒介来使用，并由 Processing、openFrameworks 等面向艺术家设计的工具支持，使计算作为艺术媒介进入更广泛的社群。计算设计，也就是使用计算生成和探索设计，从生成字体到参数化建筑，再到探索巨大工程解空间的生成设计工具，把生成方法带入设计。贯穿这些领域，计算都作为一种具有自身性质的媒介发挥作用：生成性、交互性、涌现能力，以及探索巨大可能空间的能力；艺术家和设计师正是用这些性质进行表达。

美学和创造力的形式化

第二种面向，是试图形式化美学和创造力——也就是追问：长期以来被视为最具人类特征、最抗拒形式化的美和创造力，是否能够被计算捕捉。这是这种关系触及最深问题的地方，因为美学和创造力关系到什么是人类独有的东西，而计算对它们的介入，是在探测这种独特性是否真的像它看上去那样绝对。

美学形式化有很长历史，而计算延伸了这条历史。人们一直试图寻找美的原则：比例、对称、黄金分割、音乐中的数学结构，例如自毕达哥拉斯以来把和声理解为比例。这些都暗示，审美反应具有某种可以被捕捉的形式结构。计算使这种追求得以继续：让图像或音乐令人感到美的特征能否被识别和测量？Birkhoff 的“美学度量”（aesthetic measure，1933）是一个早期形式化尝试，把美理解为秩序与复杂度之比；当代工作则用机器学习预测审美判断，并取得了一定成功。这说明审美反应至少部分可以被计算模型捕捉，尽管模型捕捉到的并不是全部。部分成功本身就很有意思：它说明美学既不是完全可以形式化的，因为模型并不能捕捉一切；也不是完全超出形式化的，因为模型确实捕捉到某些真实东西。这是计算介入所揭示出的关于审美反应本质的发现。

计算创造力——研究机器是否以及如何能够具有创造力——是更具挑衅性的介入。它追问创造力是什么，而这个问题事实证明很难精确定义；计算系统的产物是否能够真正具有创造性，还是只是看起来如此；以及机器中的创造力需要什么。Margaret Boden 的分析区分了几种创造力：组合式创造力，即熟悉观念的新组合；探索式创造力，即在某种风格或概念空间内部探索可能性；以及转换式创造力，即改变这个空间本身，这是最深也最罕见的一种。这个框架使问题可以被更精确地提出：计算系统显然能够进行组合式和探索式创造；真正深的问题是，它们能否实现转换式创造力，也就是不只是探索既有概念空间，而是真正改变这个空间。这个领域已经产生出能生成艺术、音乐和观念的系统，而且观察者会判断其产物具有创造性。这由此提出了一个问题：这种创造力是真实的，还是被归因出来的？创造力到底是过程的属性，还是观察者的判断？这个问题关系到创造力在根本上是什么。

生成式 AI 与艺术：当代对峙

第三种面向，是如今已经变得紧迫的生成式 AI 与艺术之间的对峙。能够从简单提示生成高质量图像、音乐、文本和视频的 AI 系统（§5.3，§5.5）到来之后，这种关系从专业小圈子的关切变成了每一位艺术家都要面对的问题，也以最尖锐的形式重新提出了那些古老问题：创造力、作者身份，以及人类艺术的价值。

这些问题有好几个，而且都真正困难。创造力问题强势回归：当一个扩散模型生成一张令人惊艳的图像时，它是否具有创造力？如果没有，为什么没有？毕竟如果同样的输出由人类完成，我们很可能会称之为有创造力。这会迫使我们追问创造力到底在于过程还是产物：如果创造力在于过程，那么模型的过程并不是人类创造力；如果创造力在于产物，那么产物可能已经难以区分。难以回答这个问题，恰恰暴露出我们对创造力概念本身并不清楚。作者身份问题则既实践又充满争议：AI 生成作品的作者是谁？是写提示词的人，是模型开发者，是那些作品被用于训练模型的艺术家，还是没有作者？法律维度（§9.6）在这里交汇，但更深的问题是：当作品来自一个在数百万件人类作品上训练出来的模型时，作者身份到底意味着什么？训练问题尤其尖锐，也充满痛苦：生成模型是在大量人类艺术家的作品上训练出来的，常常没有同意，也没有补偿，然后又生产出与这些艺术家竞争的作品。这造成了真实的不满，也引发了这种技术建立在挪用之上的指控。

最深的问题，是生成式 AI 对人类艺术实践和价值意味着什么。一种观点认为，AI 是一种新工具，是艺术中技术漫长历史的延续；摄影、合成器、数字工具都曾引发类似焦虑，后来也都成为艺术实践的一部分。按照这种观点，艺术家会像吸收过去技术那样吸收 AI，把它作为表达工具使用，而人类创造力仍然处于中心。另一种观点认为，这种技术不同：它不只是辅助人类创造力，而是在替代它，自动化生产原本由人类艺术家制作的文化产物，贬低人类创造性劳动，并威胁艺术家的生计以及艺术实践的意义。真相确实不确定，而且很可能因领域不同、技术发展方向不同而有所差异。这场对峙正在此刻展开：在艺术家采用或抵制工具的实践中，在创意劳动市场中，在法律中，也在文化对人类艺术价值的不断变化的判断中。这里是这种关系最具后果、也最具争议的地方；它的解决将塑造艺术本身，也将塑造计算在文化中的角色。

这种视角会改变什么

看清计算机科学与艺术的关系，会改变实践者对计算、创造力，以及技术文化后果的理解。

第一种改变，是认识到计算是一种表达媒介。实践者如果看见计算不只是构建功能系统的工具，也是一种艺术媒介——具有生成性、涌现性和交互性等自身表达特征——就会理解自己所在领域中一个单纯功能视角会遗漏的维度。计算可以被用来创造美和意义；生成艺术与计算艺术正是严肃实践这一点的领域。对于受过计算训练的人来说，这开启了其技能的一种创造性用途，而这种用途不是工程框架自然会提示的。

第二种改变，是对创造力和美学获得更经过审视的理解。实践者如果接触美学形式化和计算创造力研究，就会面对关于创造力和美是什么的问题——而计算模型的部分成功使这些问题变得具体，而不是停留在抽象层面。他们会看见，美学可以被部分形式化，但不能被完全形式化；创造力很难精确定义，而这种困难本身具有揭示性；这些最具人类特征的能力正在被计算介入探测，哪怕并未被完全捕捉。这种理解比不加批判的神秘化，或还原主义式的轻率否定，都更深入。

第三种改变，是能够清楚思考生成式 AI 与艺术。理解这场对峙的实践者，能够处理那些紧迫问题：AI 是否具有创造力？AI 生成作品的作者是谁？用人类艺术训练模型意味着什么？这种技术对人类艺术实践意味着什么？他们能带着对每个问题困难之处和真实利害关系的认识来参与讨论。这可以防止两种错误：一种是轻描淡写地说“它只是工具，没什么重要事情发生”；另一种是灾难化地说“人类艺术已经结束”。真正需要面对的是技术提出的那些真实、未决且后果重大的问题。

第四种改变，是认识到文化层面的利害关系。实践者如果看见计算与艺术的关系关乎一种文化如何评价人类创造力，关乎艺术家的生计，关乎一个社会会生产出什么样的文化产物，就会理解：他们构建的技术除了功能性影响之外，还有文化后果。生成图像和音乐的生成式 AI 不只是技术成就，也是一种文化力量。认识到这一点的实践者，会带着纯技术视角无法登记到的利害关系意识，参与这种技术的发展。

资源

基础与桥接文本

Margaret Boden 的 The Creative Mind: Myths and Mechanisms（第 2 版，Routledge，2004）是计算创造力的奠基性著作。它提供了关于创造力类型的分析——组合式、探索式、转换式——使机器创造力这个问题能够被精确提出。对于严肃思考机器是否以及如何能够具有创造力，这本书是核心文本；Boden 也是最重要地让这一问题变得可处理的哲学家。

关于计算作为艺术媒介，Processing: A Programming Handbook for Visual Designers and Artists（Reas 和 Fry，Processing 的创造者）以及更广泛的创意编程文献，是进入计算艺术创作实践的入口；关于生成艺术的书籍，例如 Matt Pearson 的 Generative Art，以及实践者的写作，则覆盖其美学和实践维度。相关历史可以在计算机艺术史著作中看到，它们记录了从早期绘图仪艺术以来数十年的实践。

关于美学形式化，文献从 Birkhoff 的 Aesthetic Measure（1933）一直延伸到当代计算美学；这个领域分散在艺术理论、心理学和计算机科学之间，计算美学综述可以作为入口。至于面对计算的艺术哲学，美学文献（连接到 §9.9）提供了概念框架。

资源 作用 类型
Boden，The Creative Mind: Myths and Mechanisms（第 2 版） 计算创造力；核心分析 深入
Reas & Fry，Processing + 创意编程文献 计算作为艺术媒介；实践入口 实践
Pearson，Generative Art 生成艺术的美学与实践 入门
Birkhoff，Aesthetic Measure / 计算美学综述 美学形式化 深入

实践与当前来源

把计算作为艺术媒介，必须通过制作来学习。创意编程工具——Processing 和 p5.js，后者是基于浏览器、易于进入的 JavaScript 版本；openFrameworks；用于交互和视觉工作的 TouchDesigner；用于音乐的 SuperCollider 以及 live-coding 环境 TidalCycles、Sonic Pi——都是制作计算艺术的手段。围绕这些工具的社群，例如创意编程社群、生成艺术场景、算法音乐社群，正是这种实践生活的地方。制作生成作品，即使只是简单作品，也是从内部理解计算作为媒介的方式。

关于生成式 AI 与艺术之间的当代对峙，相关讨论是当前的、分散的：面对 AI 的艺术家写作，包括采用它的人和抵制它的人；关于 AI 艺术的批评和理论写作；法律发展（§9.6）；以及正在进行的公共争论。这是一个必须跟踪当下讨论，而不是阅读定论的领域，因为这些问题正在此刻被解决。生成式 AI 工具本身（§5.3 和 §5.5 中的图像、音乐、文本生成器）也是相关经验的一部分——使用它们，并反思使用过程揭示了什么关于创造力和作者身份的问题，本身就是参与这些问题的一部分。

关于艺术、创造力与计算的更深哲学问题，艺术哲学和美学（连接到 §9.9）提供框架，而关于 AI 艺术哲学的专门文献正在发展中。

资源 作用 类型
Processing / p5.js / openFrameworks（免费） 创意编程工具；制作计算艺术 实践
p5.js 参考与示例（免费） 现代浏览器创意编程入口 实践
The Coding Train（免费；可选付费支持） 基于项目的创意编程视频 辅助
OpenProcessing（免费；Plus 选项付费） 创意编程社群与 sketch 档案 实践
SuperCollider / TidalCycles / Sonic Pi（免费） 算法音乐与 live coding 音乐 实践
生成式 AI 工具（§5.3，§5.5） 当代媒介，需要直接体验 实践
当代 AI 艺术讨论（持续发展中） 正在发生的对峙 参考

陷阱

陷阱 为什么会误导 更好的回应
把计算艺术斥为不是真正的艺术 这种观点认为用计算制作的艺术——生成艺术、算法音乐、AI 辅助作品——不是真正的艺术，因为真正的艺术需要人手，或者因为是计算机完成了工作。这误解了艺术，也误解了媒介。计算艺术包含真实的艺术选择：生成系统的设计、输出的选择与策展、对媒介属性的表达性使用。否定它通常反映的是对这种媒介的不熟悉，而不是一种站得住脚的美学立场。每一种新媒介都曾面对这种否定，例如摄影、电子音乐，而这些否定后来都被证明是错误的。 把计算艺术当作一种具有自身属性和自身艺术实践的媒介来对待。在生成系统中工作的艺术家做出了真实的创造性选择，而这种媒介的性质——生成性、涌现性、交互性——本身就是表达资源。应根据作品的美学价值来判断作品，而不是否定媒介。艺术史就是新媒介不断被否定、随后被纳入的历史。
以为创造力和美学要么完全可形式化，要么完全无法形式化 这里有两个相反错误：一种还原论观点认为，美和创造力只是等待被发现的可计算函数；另一种神秘主义观点认为，它们完全超出任何形式理解。来自计算美学和计算创造力的证据位于二者之间：这些能力可以被部分形式化，模型确实捕捉到某些真实东西；但它们并不能完全被形式化，模型也没有捕捉一切。部分成功本身才是有意思的发现。 保持证据所支持的中间立场：美学和创造力具有计算可以部分捕捉的形式结构，同时也存在当前形式化尚未触及的剩余部分。审美判断计算模型的部分成功，以及计算系统真实但有限的创造力，都是关于这些能力性质的发现——它们既不是完全机械的，也不是完全神秘的；真正有意思的问题是，究竟哪些东西被捕捉到了，哪些没有。
过早裁定 AI 是否具有创造力 有些人把生成式 AI 是否具有创造力看作显然为是，因为它生产出看起来有创造力的作品；另一些人则认为显然为否，因为它只是统计，不能真正创造。双方都把一个真正困难、并且揭示我们对创造力概念并不清楚的问题，当成已经解决的问题。这个问题的困难正是重点：它暴露出我们并没有关于创造力的清晰解释。 把这个问题当作真正困难且具有揭示性的事情来处理。Boden 关于组合式、探索式、转换式创造力的区分，使这个问题可以被更精确地提出；AI 是否实现了转换式创造力，仍是开放且深刻的问题。回答“AI 是否具有创造力”之所以困难，本身就说明创造力概念并不清楚；认真处理这种困难，比在任一方向上给出自信答案更有价值。
忽视对人类艺术家的利害关系 关于 AI 与艺术的问题很容易被抽象讨论——创造力、作者身份——却忽视对人类艺术家的具体利害关系。模型用艺术家的作品训练，这项技术可能威胁他们的生计，而他们关于作品未经同意和补偿被使用的不满是真实的。把这种关系当作纯粹的思想谜题，会忽视真实的人正在受到影响，有时甚至受到伤害。 把思想问题与人的利害关系放在一起看。模型在没有同意和补偿的情况下使用艺术家的作品训练，这是一种真实的不满；艺术家生计受到威胁也是真实的；自动化创意生产的文化后果也很重要。负责任地处理这种关系，意味着关注受影响者面对的具体后果，而不只是讨论关于创造力和作者身份的抽象问题。