Generative machine learning models' increasing prevalence and capacities are transforming creative processes. We identify two commonly voiced threats to professional artists: i) Barrier Reduction, i.e. AI enabling laypeople to engage in creative work; and ii) Autonomous Creativity, i.e. a reduced need for human input in (semi-) autonomous AI systems. Has AI already leveled the playing field between professionals and laypeople? We address this question by experimentally comparing 50 professional artists and a demographically matched sample of laypeople. To this end, we designed two tasks that approximate artistic practice in both faithful and creative image creation: replicating a reference image, and moving as far away as possible from it. We developed a bespoke platform where participants used a modern text-to-image model to complete both tasks. Artists, on average, produced more faithful and creative outputs than their lay counterparts, highlighting the continued value of professional expertise --- even within the confined space of generative AI itself. We also explored how well an exemplary vision-capable large language model (GPT-4o) would complete the same tasks. It performed on par in copying and slightly better on average than artists in the creative task, although not above the top performers in either human group. These results highlight the importance of integrating artistic skills with AI training to prepare artists and other professionals for a technologically evolving landscape. We see a potential in collaborative synergy with generative AI, which could reshape creative industries and education in the arts --- just as other technologies like photography have done before.

Philosophical traditions linking vision and knowledge have long shaped our conceptions of humanity. Yet these are now challenged by neuroscience’s revelations about the bodily underpinnings of imagination, as well as by AI’s capacity to generate synthetic images from textual prompts. If vision once epitomized a direct route to knowledge, present-day technologies entangle sight with computation, prompting a thorough re-examination of originality, authorship, and the nature of understanding. This research investigates how the interplay between human cognition and AI-driven imagery reshapes both personal and collective imagination, thereby reframing centuries-old assumptions about visual perception as a privileged conduit to meaning. By bridging “individual” tastes with vast collective archives, generative AI unsettles how cultural artifacts are created, shared, and interpreted, fuelling philosophical inquiry into how our imaginative faculties may be subtly reconfigured through machine-driven “cognitive hacking.” Indeed, text-based interactions with algorithmic latent spaces challenge conventional boundaries of novelty, often blurring the line between genuinely new vision and recycled pattern-based recombinations. In this evolving synergy between human and machine, biases embedded in training data become potent creative constraints, raising ethical and epistemic questions around shifting norms of authorship and artistic intent. Ultimately, as neural models encapsulate and re-model culture, AI-driven imagery emerges not merely as a practical tool but as a transformative force that alters how we see, think, and know. By reordering the transmission of cultural meaning and recasting vision’s once-straightforward link to truth, generative AI compels us to reconsider the conditions under which knowledge is produced, unveiling fresh possibilities and unforeseen perils for creativity and perception alike.

AI has become an umbrella term for a myriad of data driven technologies and forms of generative media production. As such, they have revealed forms of human exceptionalism to be - more than anything - narratives, framings, and constructions, which have real, material implications. As such, notions of creativity, intelligence, art, science and so on are prone to be reconfigured and performatively brought forth by human machine collaboration. Drawing on the experiences and collaborations of AI: Ancestral Immediacies, my contribution will reflect on artistic practice as media theory in times of AI.

In an era dominated by simulated AI, the tangible and embodied dimensions of intelligence—rooted in the physical body, time, and materiality—are increasingly overlooked or forgotten. This talk focuses on the intersection of painting, computer graphics, and robotics as tools for exploring embodied knowledge and creativity in the age of AI. Through the presentation of selected art projects that merge biologically inspired robotic systems, deterministic robotic processes, and human-machine interactive painting methods, the talk will investigate new forms of hybrid human-machine creativity and the practical implications of this convergence. Emphasizing the growing divide between digital (simulated) and analog (physical) realms, it will question how this shift impacts a society becoming increasingly disconnected from embodied experience. The talk will feature artworks from the Embodied Agent in Contemporary Art project as a case study, showcasing how these emerging technologies challenge and redefine the very nature of creativity.

As artificial intelligence continues to reshape creative industries, the relationship between human and machine creativity demands deeper exploration. This presentation will examine key questions surrounding AI’s role in augmenting creative processes, drawing on insights from my research and artistic practice. My work spans AI-human collaboration in solo and team creativity, the neuroscience of creativity, and neurodesign—an approach that integrates cognitive science into AI system design.A central focus will be on how AI can enhance human creativity, not simply by automating tasks but by facilitating cognitive states such as Flow, a critical condition for peak creative performance. Through performance-based neuro-art, I use wearables to stream neurophysiological data, which is classified by AI and transformed into audio-visual outputs designed to encourage and sustain Flow states. This approach highlights how computational systems can dynamically interact with human cognition to expand creative potential. Beyond individual creativity, the talk will briefly explore how AI is reshaping artistic practice and cultural transmission, along with key challenges like deskilling, model collapse, and shifts in creative expertise. Finally, I will discuss how human behavior and cognitive insights can inform the development of future AI tools, ensuring that they amplify rather than diminish the richness of human creative expression. Through an interdisciplinary lens, this presentation will offer new perspectives on the evolving synergy between human ingenuity and machine intelligence.

Advancements in artificial intelligence (AI) have significantly transformed the art industry by automating creative processes and reshaping art production and perception. AI-powered tools like Stable Diffusion, Midjourney, Adobe Firefly, and DALL-E generate high-quality images by processing vast datasets, sparking debates about the authenticity and artistic value of AI-generated art. While proponents argue that AI creations exhibit creativity and deserve legal recognition, critics question human authorship and ethical implications. Artists using the NMKD Stable Diffusion GUI with human-in-the-loop techniques to generate AI variants of their existing artworks found the process enriching and collaborative. They appreciated AI's ability to help them explore new perspectives while maintaining their creative essence, preferring AI variants over entirely AI-generated pieces. This preference underscores the value of retaining creative control and personal connection to their art. Our work emphasizes the importance of a harmonious partnership between AI technology and human artistic expression, highlighting ethical standards and broader implications. It discusses legal challenges such as copyright infringement and the need for responsible use of AI in art. The integration of AI raises questions about the future of human artists and the value of human creativity, suggesting that AI should complement rather than replace human creativity. In conclusion, while AI offers exciting possibilities for artistic expression and innovation, its integration must be approached with caution and responsibility to ensure a positive societal impact.

Novelty is a key component of creativity, and it appears that (at least superficially) generative AI systems can produce novel outputs. However, critics of generative AI have noted the ‘generic’ or ‘samey’ nature of AI art. This raises the question of whether generative AI systems are really capable of producing adequate novelty to be considered truly creative. This paper argues that popular generative AI systems do indeed have a novelty problem. The perceived banal nature of AI outputs can be explained by a lack of ‘originality’. Originality is often seen as synonymous with novelty, but it can be distinguished as salient novelty (see Gaut). This paper puts forward an account of what would count as salient newness in AI images. Given the vagueness of ‘salience’ in this definition, we can utilise work by Sibley on originality to uncover three criteria for salience in novelty: it must be relevant; it must be attributable to the creator; and it must exhibit a moderate to significant degree of variation from prior works. In considering the future of machine creativity, I suggest an increased focus on this third criteria: mechanisms of variation. The requirement for variation in Sibley’s account not only thickens the account of originality but also meshes with the evolutionary account of creativity; however, variation in evolutionary processes is often minimal in the short term. For the more extreme or transformative kinds of creativity, we need larger variances. This paper therefore discusses the need for significant variation for originality in computational creativity.

As artificial intelligence technologies reshape how language is structured, circulated, and commodified, they emphasize the fact that language is not merely a medium of communication but a primary instrument through which power is manifested, exercised and reinforced. SolidGoldMagikarp: Artistic Interventions in the Political Latent Space of Large Language Models paper presents artistic research into the political, epistemological, and ideological dimensions of Large Language Models (LLMs), through experimental misuse, critical coding, and speculative hacking. I approach LLMs as more than tools – they are complex, opaque, techno-social systems that encode worldviews, political ideologies, and sociotechnical imaginaries, often invisibly, subtly and seamlessly. The project, rooted in a critical art practice, engages with models like ChatGPT, Claude, Gemini, DeepSeek and YandexGPT as both media and material. I also work with text2image and text-to-3D diffusion models, based on LLMs. By pushing these systems beyond their intended use – generating images from abstract or politically charged prompts, forcing ideological consistency through recursive questioning, and testing censorship boundaries – I aim to map the latent space not only as a technical construct, but as a field of power. In this way, I examine how language is fragmented, tokenized, and reformatted for profit, control, and ideology.

Generative AI is entering contemporary artistic practices. This talk explores how artists are harnessing these types of neural networks for creative expression across diverse media practices, including image, video, text, and form generation. Drawing on practice-based research, we examine both the potential and limitations of current generative AI technologies. However, we argue the importance of open small AI models hinges on the artist's creation of assemblages—unique configurations using AI components to realize their distinct voice and artistic visions. We also highlight the key importance of the open small AI models for artistic exploration in media art.

Generative AI has swiftly and profoundly transformed the creative art landscape, rendering traditional tools seemingly obsolete and causing significant shifts in how artists approach their craft. This rapid evolution has led to widespread misconceptions, particularly among those who closely associate artistic value with the tools used, leading some to believe that AI might replace aspects traditionally considered exclusive to human creativity. However, generative AI fundamentally represents a radically different category of artistic tool—one capable of significantly expanding an artist's expressive potential rather than replacing creative talent itself. The confusion and anxiety around AI's role in art primarily arise from the current limitations in interacting with generative models, which predominantly rely on simplistic text prompts. The limited expressive capacity of text alone underscores the urgent necessity for more sophisticated interaction mechanisms, especially for detailed creative vision communication in image and video generation. Future developments must focus on creating interactive, multimodal tools that facilitate deeper and richer communication between artists and AI models. These tools will need to incorporate a thorough understanding of physical principles, human contexts, and three-dimensional spatial awareness. Additionally, artists will require innovative methods for capturing and integrating real-world contexts with AI-generated content, ensuring seamless blending between generative outputs and captured reality. This talk will explore recent advancements in interactive generative AI tools, highlighting promising directions toward enhanced, intuitive, and artistically empowering human-AI collaboration.

Can generative AI be genuinely creative, or does it merely remix existing human ideas? While AlphaGo revolutionized the game of Go with strategies previously unimagined—and now adopted by expert players—the question remains open for more open-ended domains like art. In this talk, we present a framework for studying creativity and cultural evolution in the visual arts through generative AI. We introduce a large-scale dataset of 1,114 artists spanning the 1400s to 2000s, from which we derive “style embeddings” using textual inversion on Stable Diffusion’s CLIP component. Our analysis reveals that artists tend to cluster by historical era, while the convex hull of styles expands over time, suggesting the continual emergence of new stylistic directions. We further model each artist’s style as a linear combination of their predecessors, showing that many styles can be explained by a small subset of influential forerunners. However, we argue that numerous potentially compatible artistic concepts remain unexplored, not because they are fundamentally incompatible, but because of historical, social, and cognitive biases. Generative AI, despite its sweeping training data, inherits these biases from human culture. Inspired by research in algorithmic scientific discovery, we propose a system designed to counteract these biases in order to unveil “culturally inaccessible” concept combinations—such as Renaissance-style airplanes—that lie outside the historical record. Our results highlight the potential for AI to transcend cultural constraints, offering new avenues for examining and shaping the future of artistic innovation.

Theorists of creativity often treat intention as a necessary condition for creativity. On this view, an agent must intentionally produce outputs that are both novel and useful in order to count as creative. When such outputs arise accidentally, these theorists claim that only pseudo-creativity has occurred. Since generative AI produces novel and useful outputs without intention, some argue it is, at most, pseudo-creative. In this presentation, we challenge that view and argue that the intention criterion should be abandoned. We offer three arguments in support of this claim. First, even in the case of human creativity, the intention requirement is unacceptably vague. Second, creativity has long been attributed to sources that lack intentional agency. And third, both general and expert usage of the term creativity (and related terms) has evolved in response to the rise of generative AI. Rather than dismissing this usage for failing to match existing definitions, we suggest it indicates that we need to revise the concept of creativity itself—specifically, by dropping intentional agency as a necessary condition.

Standard theories of creativity possess two features. First, they are monistic: for them, questions of the form: “Is x creative?” or “Does creativity matter?” are to be answered by appealing to a unique, fundamental conception of creativity. By this, I don’t mean that they don’t recognize several notions of creativity; most of them do. Rather, I mean that they are committed to the claim that there cannot more than one fundamental notion of creativity. Second, their conception of creativity is kainotic (from the Greek kainos, innovation). That is, they assume that the fundamental notion of creativity, the one in virtue of which all the others can be defined, must be understood in terms of unprecedented novelty. Margaret Boden’s canonical definition of creativity, for instance, clearly displays these two commitments: her notion of psychological creativity is both fundamental and defined in terms of psychological processes that, as a type, are responsible for the generation of unprecedented novelty. In my presentation, I will argue that we have both conceptual and empirical reasons to reject these two aspects of standard theories of creativity. A better theory of creativity should be a pluralistic one, where different fundamental notions of creativity collaborate with each other. In particular, I will distinguish between kainotic and non-kainotic fundamental conceptions of creativity and will sketch an account of how these could be combined to offer us a better account of what is creativity and of why it matters.