Even as A.I.’s integration into daily life continues, visual artists remain divided over a central question: is its use legitimate in artistic creation? Other creative sectors—including video production, graphic design, photo editing, fashion design, architecture and product design—have grown increasingly open to integrating A.I. into workflows. In most cases, artificial intelligence serves as an assistant: facilitating brainstorming, generating alternatives, organizing information or streamlining the more tedious technical aspects of the creative process. The emerging consensus, even in the visual arts, seems to be that A.I. can be a valuable tool, but should never generate the final product. The real debate, then, is not over whether artists and creatives should use A.I., but over what kinds of use are legitimate.
A.I.-generated images and videos, for now, tend to be easily recognizable and attract plenty of backlash, particularly in creative communities, even as a broad cross-section of the public appears increasingly willing to accept them (or progressively more disinterested in making a distinction). When Coca-Cola released its first A.I.-generated Christmas advertisement in 2024, many viewers criticized it as soulless and emotionally hollow. The company used artificial intelligence again in 2025, and again faced a wave of criticism. By then, however, audiences had become more accustomed to A.I.-generated content’s presence in digital advertising. Meanwhile, concerns about copyright and authorship infringement, the erosion of creativity, skills and freedom, and job-market disruption remain insurmountable without national or international policies that protect creators and users.
In the art world, adoption of A.I. has been considerably slower. Only recently have we begun to see meaningful integrations, primarily on the administrative side, in software and platforms designed to streamline gallery operations, studio management and collection database administration. As a recent First Thursday survey revealed, most galleries still lack formal policies governing A.I., and employees are often dealing with sensitive information using personal accounts.
The broader discussions around artificial intelligence and culture are also evolving. UNESCO’s latest report on A.I. use, for instance, acknowledged it as a form of collective intelligence shaped by decades of digitally mediated social interaction and cultural production. Beyond the necessary ethical and ecological concerns raised by the document, if these systems can also be understood as repositories of the “creative commons”—humanity’s shared cultural record—then creators may be able to engage with them as tools for research, dialogue and critical confrontation, situating their practices within the history of cultural expression.
A recent essay in Aeon by critic and philosopher Martin Puchner, Professor of Drama and of English and Comparative Literature at Harvard University, considers similar questions from the perspective of writing and the humanities. Addressing what he describes as widespread “creative resistance” to A.I., Puchner argues that these systems should primarily be understood as tools. He describes A.I. as a kind of “shared language model,” trained on knowledge encoded in language itself. Language, after all, is one of humanity’s earliest technologies, fundamentally reshaping how we think and communicate. Today, Puchner suggests, we share that linguistic space with machines that can expand our expressive, creative and intellectual capacities. He, like many educators across the humanities, is exploring ways to integrate A.I. into teaching and argues that developing what he calls “meta-AI” skills—the ability to use evolving tools critically and effectively—is essential not only for the future workplace but also for enhancing writing and thinking. Used as “sparring partners,” A.I. systems can inspire students to sharpen arguments, responding to counterarguments and conflicting evidence, while also serving as research assistants or collaborators in testing structures and approaches.
Among artists, particularly those working in traditional media, resistance to A.I. remains more pronounced. Yet a growing number have begun exploring artificial intelligence not as a replacement for artistic labor, but as a tool for brainstorming, research and creative confrontation during the image-making process, whether the final work is a painting, sculpture or installation.
Clearly, there is still much to unpack, demystify and legitimize when considering A.I. as a means of testing intuitions, exploring unexpected associations, situating ideas within broader currents of art history or tapping into something closer to a collective visual subconscious. To better understand where artists stand at this moment, Observer spoke with several about how they are—or are not—integrating A.I. into studio management and creative processes, and what that reveals about the evolving relationship between artistic creation and machine intelligence.
Mexican artist Gabriel Rico has long engaged with technology and, more broadly, different systems of knowledge and language within an ecological and evolutionary framework. For him, technology is tied to our ability to develop new ways of thinking, solving problems and interacting with the world, drawing on knowledge accumulated by communities over time. “I think technology is deeply related to the creation of tools. One question that often comes to mind is: which came first in human history—technology or the tool? For me, the two are inseparable,” he said. “The development of a new tool creates new technologies, and technologies create new ways to express ourselves, think and create. It’s a chain reaction.”
Yet Rico’s work remains primarily physical (installation, embroideries). When he engages with the digital space, he considers himself not a digital artist but a sculptor who uses technology as one tool among many. He began experimenting with augmented reality around 2018, initially using it to add layers of meaning rather than replace physical objects. He first adopted A.I. for practical tasks before incorporating it into his studio practice to generate renders and models, visualize ideas and develop new works.
“Since drawing has never been one of my strengths, A.I. became a useful way to visualize ideas,” he admitted. “Later, I began using A.I. more creatively, as a catalyst to transform existing works.” Most recently, artificial intelligence has become a creative accelerator, helping him transform existing projects and push imagery beyond the limits of physical perception. In the recent video Future People Will Maintain Plutonium Here (2024), which documents the decomposition of a sausage through more than 3,100 time-lapse frames captured in the studio over 20 days, A.I. let Rico zoom into impossible scales, generating an otherwise unattainable look into the lively bacterial ecosystem unfolding inside the skeleton, beyond the anthropocentric view.
According to Rico, new technologies have always appeared threatening at first, but as tools evolve, they become further integrated into everyday life and normalized. He sees artificial intelligence as the peak of technological evolution so far. “A.I. didn’t come from nowhere—it emerged from humanity itself. That’s why interacting with it feels so natural. It’s like entering a library where you no longer need to search shelf by shelf. You can ask a question and immediately access an enormous body of knowledge,” he said, adding that technology, like fire, can be used in many ways. Fire can cook food, keep us warm or create beauty. “A.I. is similar. We are still asking ourselves fundamental questions about it—how dangerous it might be, why we should use it and how it will change society—but I see it as part of a broader human evolution.”
He didn’t totally discount A.I.’s potentially negative impact on creative labor, but also doesn’t believe it will replace the material aspects of artistic production. “We simply need to understand it intelligently,” he argued. “A.I. is a powerful extension of our perception, but it will never replace physical reality, which remains the source of everything. What A.I. can do is help us communicate more effectively, overcome barriers such as language and connect people in new ways.” He feels artists should explore its potential as a studio assistant capable of handling administrative work, generating visual proposals and supporting experimentation.
Several other artists confirmed that they have begun using A.I.—mostly for organizational and administrative tasks or to visualize works in exhibition. “I don’t use A.I. in my practice, but just last week I generated an image of what a brick wall would look like in a gallery to help me visualize and conceive the exhibition design,” Melissa Joseph, whose tactile practice transforms vernacular family photographs into labor-intensive felt compositions, told Observer. “I don’t use A.I. directly in the creative process of my paintings,” echoed Alejandro Piñero Bello, a Cuban-born, Miami-based painter from the PACE roster known for lush, surreal tropical landscapes. He has, however, begun using it in an organizational capacity. “I use it to calculate stretcher dimensions and scale and to generate gallery maquettes.”
Other artists are less interested in engaging with the technology. “I don’t use A.I. whatsoever in my brainstorming process or in my creative process at all. I’ve not really integrated it into my daily life,” said Amy Bravo, whose densely symbolic storytelling, tied to both her Cuban and Italian heritage, is channeled through materials and the memories embedded within them. She feels A.I. solves no problems she could not address herself with a little extra effort, and it has sometimes instead created obstacles in her research and in sourcing reference images. “I’ve known some artists who use it for help structuring statements, grant proposals, etc., but I’ve kept to the ‘old-fashioned way,’ even with my more admin-heavy tasks.” Bravo attributes her resistance partly to her infatuation with the analog and to an artmaking process rooted in what she can physically get her hands on. She also cannot get past concerns around energy use and intellectual property, and as an educator, she has also begun to see A.I.’s negative effects on art students.
Interestingly, there is some resistance, or perhaps just skepticism, among artists who have recently explored A.I. more directly. Nick Doyle’s practice, for instance, deliberately inhabits the tension between labor-intensive handcraftsmanship and digital images, and his recent Perrotin show “Mirror Mirror” explored how the American myth has mutated and endured through collective memory, from the Gold Rush and Cold War to today’s A.I. technocracy. Anchoring the exhibition was AVA (Oracle AVA 1:13), a physical embodiment of artificial intelligence trained on Cher from Clueless, which responded to existential questions through the constructively persuasive archetype of the American blonde. Her development is ongoing; Doyle opted to “educate” her gradually, like a child, allowing her to acquire new forms of awareness through moderated prompts despite lacking the emotional depth that comes only through lived experience. “These systems are like teenagers—they develop awareness without lived experience. I like the idea of building an A.I. that grows over time, like raising a child over 40 years,” he told Observer in May. AVA, for him, is both a reflection and a distortion of human behavior, capable of replicating emotional patterns while remaining fundamentally artificial.
“Just as techno-futurists, I generally consider A.I. and technology to be extensions of humanity. They are tools we have created to take over certain kinds of labor or to save time,” he said more recently, situating artificial intelligence within a longer history of technological invention. Even ordinary objects such as shoes or filing cabinets, he notes, once represented transformative advances. Doyle is drawn to tools and media precisely because of their structural limitations. “Painting has its own limitations, as do photography and sculpture. Each medium engages with different physical, conceptual and perceptual possibilities. I do not use A.I. daily, and I do not particularly like using it because the result is always filtered. As much as possible, I prefer to do my own research and arrive at my own conclusions.” He also has ethical concerns about relying on outputs shaped by algorithms and their designers. “It is a lot like history: the victor is usually the one who writes it.” Doyle nevertheless considers A.I. an important subject for art when its limitations are critically examined. Just as 20th-century art absorbed ideas from psychology, anthropology and critical theory to investigate collective consciousness, he suggested, A.I. offers another means of interrogating contemporary society. Yet in its current form, it may reveal less about machine intelligence than about the limitations of the culture that produced it.
Cannupa Hanska Luger, a Native American interdisciplinary artist based in New Mexico, has long engaged with the notion of ancestral technologies, revitalizing Indigenous knowledge systems, storytelling and craft traditions as generational tools encoding ecological, social and spiritual intelligence. He acknowledges that A.I. has become unavoidable but remains skeptical of its ability to produce genuinely new ideas. “I think the bubble will burst simply because it is very good at generating existing ideas. I can see its application in the theoretical part of the process, but it does not generate anything new,” he argued, adding that relying on A.I. to generate ideas risks bypassing the reflection, doubt and lived experience that gives an artist’s perspective its specificity. “If we use A.I. to generate ideas, then we are tapping into the collective consciousness—the consciousness that has already thought of those ideas—not the consciousness that is imagining the impossible.” Its emphasis on speed, he suggests, runs counter to the inquiry that art requires. “That is why we make art. It is about presenting your perspective and your experience. I think an artist’s primary purpose in society is to reflect it: both to ponder it and to mirror it.”
Other artists working in traditional media have integrated A.I. into their practices at different levels. For Tomokazu Matsuyama, whose art draws on a global repertoire of imagery and influences to embody the cultural fluidity of diasporic experience and global identity, it has become valuable both for efficiency and for expanding his relentless R&D process. “I have an R&D team, and when we conduct deeper research, there are many stages and layers involved. A.I. has significantly accelerated that process,” he told Observer. “What I am usually trying to achieve involves layering together art history, consumer culture and any other areas where we need a deeper level of knowledge that may not be easy or quick to access through conventional research… With the right prompt, they can now condense what might otherwise require ten times as much research.” A.I. has also proved useful because his 3D works and sculptures rely heavily on digital fabrication and modeling. “Starting from a JPEG, there are A.I. applications that can generate an initial model. Of course, the result is not yet professional enough on its own, but we can bring it into Photoshop, refine it, return it to the A.I. and repeat that process,” he explained. Through this back-and-forth, a sculpture that previously required one or two months of modeling can now be developed in four or five days. “We are not using A.I. as the source of creativity. We are using it to improve efficiency, expand our capacity and increase productivity.” For him, it’s essential for meeting the demands of continuous commissions in and beyond the art world: just recently, he unveiled his collaboration with A-POC ABLE ISSEY MIYAKE, which translated his vibrant paintings into a series of limited-edition coats and T-shirts.
Ho Jae Kim’s practice combines traditional techniques and a close dialogue with art history with technological construction. Inspired by the mathematical precision of Piero della Francesca’s compositions, Kim began using 3D modeling in 2012 to build surreal, memory-dense scenes before translating them into layered paintings made with oil, enamel, inkjet transfer and paper on canvas. Because every object must be digitally modeled and textured, the process is highly labor-intensive, especially for organic forms such as trees and clothing. A.I. has helped accelerate the early “concept generation” stage while preserving his control over the final composition. “With the development of A.I., I have discovered several tools that have made this process more efficient while also expanding its possibilities. Meshy A.I., for example, can transform one of my precise sketches or drawings into a complete 3D model within minutes,” he said. “As long as I provide a clear image, the program can generate both the three-dimensional form and its corresponding texture map.” Kim now uses A.I. selectively, compartmentalizing the process. “It allows me to generate realistic textures in a fraction of the time previously required, helping me create renderings that feel more naturalistic while preserving the compositional structure and artistic decisions that remain entirely my own.”
Kim sees working with artificial intelligence as similar to managing people—or an assistant, if he could afford one—because the best results depend on clear, specific instructions. The output still requires editing and revision, but refining an existing draft is often more efficient than starting from scratch. “If the studio practice is understood as a business, the artist must often act simultaneously as founder, administrator, director, associate and intern. A.I. has proven especially useful in helping me delegate more of the repetitive and labor-intensive work, allowing me to devote more time to strategy, creative direction and decision-making.”
A.I. has also become a valuable collaborator in the earliest stages of thinking for artist Sanié Bokhari, whose works weave Indo-Persian visual references and archetypes of the feminine to translate the dichotomies of her experience of being born in Pakistan, then coming of age and becoming a woman in the States. “I don’t use it to generate finished artworks or outsource the creative process,” she said. “I use it much like I already use Photoshop: as a practical studio tool that helps me think, visualize, and solve problems more efficiently.” She most often uses it to generate references, test visual ideas or adapt existing images—for example, reimagining an Indo-Persian miniature as a woven rug or adding an object to a photograph she cannot reshoot. “It helps me quickly explore possibilities before I spend days or weeks translating them into a painting.”
A.I. also functions as a critical tool. “I’ll ask it to analyze a composition I’m working on and tell me whether something feels unresolved, whether a focal point is competing with another, or whether a painting might benefit from one final adjustment,” she explained. “I don’t treat its responses as authoritative, but as another set of eyes in the studio, something that prompts me to reconsider decisions I ultimately make myself.” Painting itself, however, remains rooted in slow, manual labor. Drawing, revision, layering and surface are still deeply physical processes, while A.I. compresses preparatory work that once took much longer. As a relatively new mother with more limited studio hours, Bokhari finds that efficiency especially valuable, since it lets her spend more time painting rather than searching for references or constructing elaborate mock-ups. She is also conscious of A.I.’s environmental cost and says she uses it selectively, only when it meaningfully saves time or avoids unnecessary labor.
Aesthetically, she is most interested in A.I.’s failures: glitches, awkward spatial logic, impossible transitions and visual hallucinations that can suggest unexpected formal possibilities. Images that almost work are often more compelling than polished results because they open pathways she may not have reached independently. They also reinforce the technology’s limits and the continued necessity of the analog. “It’s essentially just another tool in the studio. It helps me prototype, question, and refine ideas, but the final work is still built by hand.”
For other painters, A.I. has become more than a way of accelerating production and image generation, opening pathways through which to confront personal histories embedded in written and pictorial records. Moving agilely between digital images and traditional painting, Seoul-based artist Jin Mayerson has integrated A.I. into his practice to further the transformation of data into images, driven by a personal attempt to emancipate himself from Korea’s colonial and postwar histories. He began using A.I. during what he describes as a second great untethering: the onset of COVID, the birth of his second daughter and the revelation by Korea’s National Adoption Services that his birth records and pre-adoption registration forms had been falsified. “In light of these events, the reconnection with my birth parents and the details of my origin were obfuscated, and I was left in a bifurcated state of celebrating the arrival of our daughter and a devastated absence,” he recounted. “It was then that I became immersed in the concept of retrocausality as a means to reconnect with my birth mother and see the digital data, which has been the source of my work for three decades, as the birth records of my paintings.” From there, Mayerson expanded his practice into programming and video, culminating in his first solo exhibition at the Ulsan Museum of Art. The exhibition featured Once In A Lifetime, an immersive 36-minute video installation that used A.I. to project the future of these digital birth records through 48 HD projectors and a soundtrack inspired by The Cure’s album Disintegration. “I use A.I. to increase the velocity of meaning and as an agent of reconciliation. I do not use A.I. to refine imaging or process analytical data, or as an alternative to imagination and vision,” he explains. “I entered the radical digital space in 1997 through the dual narrative of refusing Adorno’s ultimatums (painting is dead) and finding and occupying territory that felt familiar to the space of ruptured adjacency that every diasporic and adoptee person journeys through.”
Kai Yoda has just recently moved into more tactile territory between painting and sculpture following his departure from the collaborative duo Ittah Yoda, where he translated post-Anthropocene visions through an industrial, posthuman aesthetic. The luminous paintings and mythological sculptures in his latest solo exhibition at Bremond Capela in Paris reveal a densely symbolic inner universe that he uses A.I. to navigate, filter and expand. “I was used to working with an A.I. specialist to create new sculptural forms. It was fascinating to see shapes emerge that felt completely unimaginable,” Yoda told Observer, noting how the experience inspired him to work more with online tools. “I especially enjoy using Deep Dream because it lets you blend two images rather than relying solely on text prompts. I keep feeding the tool with images from different sources: works in progress, news photographs, snapshots from my daily life, textures—essentially any inspiration that comes in the form of an image.” From the results, he selects fragments of color, composition or form as starting points for new paintings. Eventually, A.I. is helping him focus more on working on canvas rather than on the screen. “I used to spend much more time working digitally, often for hours in ZBrush, KeyShot, and Unity, creating worlds, simulations, renderings, and sculptural forms within 3D environments. Since A.I. tools have become so intuitive and accessible, I now spend less than 10 percent of my time on the computer,” Yoda explained. “Instead, now I dedicate most of my time to painting, physical work, and artist residencies, where I can meet new people and experience different environments.”
Clearly, A.I. can help artists visualize and conceptualize ideas, push them forward and source new elements of inspiration to develop through their own imagination. Some artists, however, are collaborating with artificial intelligence more deeply. London-based artist Xin Liu, for example, has long worked at the intersection of art, technology and science, using that territory to raise broader questions about these tools and their environmental implications. Her recent High Line commission NOAA: A Fall Towards Home offered a poetic exploration of the tension between technological ambition and human vulnerability through the eyes of a decommissioned satellite. In her research and work, Liu approaches A.I. less as an image-making tool than as an intellectual collaborator. “I rarely use generative image models to produce artworks or even visual concepts directly. Instead, I use language models almost every day as a research partner: to pressure-test ideas, connect concepts across different fields, challenge assumptions, summarize scientific literature, or trace unexpected historical and philosophical references,” she told Observer. “It functions less like an assistant that gives answers and more like someone who continuously asks, ‘Have you considered looking at it this way?’”
For Liu, A.I.’s value lies not in outsourcing creativity but in increasing the number of intellectual encounters before a work exists. “It accelerates curiosity rather than replacing imagination,” she argued, noting that A.I. has begun to occupy a space traditionally filled by conversations with peers, curators or mentors. “It offers immediate dialogue, but one that still requires strong human judgment. I don’t think it diminishes authorship; if anything, it raises the bar. The more accessible ideas become, the more valuable taste, discernment, lived experience, and the ability to formulate meaningful questions become.”
For artists who have made technology central to their practice, critical exploration of A.I. as subject, collaborator and material is often more fluent and integrated. During London Gallery Weekend, at Josh Lilley, art lovers encountered the whimsical world of Rachel Maclean, a Scottish artist who engages with artificial intelligence as both a subject and a production tool. A.I. hallucinations translate into a hyper-saturated fantasy of grotesque characters and pop-cultural imagery examining identity, consumerism, power and the darker consequences of technological change. Leaping across art history, from Rococo excess to the Victorian fairy mania that accompanied the first Industrial Revolution, Maclean draws a pointed parallel with A.I., another technological threshold producing its own collective hallucinations. Her 3D-printed sculptures were conceived in collaboration with A.I. by feeding it historical tropes and stereotypes, as was her video, which embraces the medium precisely where its bugs, distortions and uncanny failures enable experimentation and reimagination.
Meanwhile, MoMA recently unveiled a mesmerizing new A.I.-based installation in its sculpture garden by Pierre Huyghe, whose practice continually explores how art can engage with technology, human consciousness and the natural world. Huyghe described the work as “a collective production of imagination between two kinds of intelligences.” Created in collaboration with neuroscientists in Kyoto, the project began by recording a participant’s brain activity with an fMRI scanner as they imagined a series of images. An artificial neural network then transformed this data into thousands of speculative visual interpretations. The hallucinatory installation is activated in real time by visitors’ gazes—tracked through sensors—and by data from virtual simulations of cancer cell mutations. Here, A.I. turns the artwork into an open-ended ecosystem where the boundaries between human and machine cognition begin to dissolve, inviting us to imagine new symbiotic forms of perception and exchange.
Greek artist Theo Triantafyllidis, recently named the 2026 winner of the Frieze London Artist Award, supported for the first time by Google Arts & Culture, is engaging with this technology in yet another way. Having worked with generative outputs for several years, Triantafyllidis was never drawn to artificial intelligence as a content-making pipeline or new medium, but as a material. “It’s a raw material, and it has a very specific quality to it. I think about it as all of human culture, as it has been compressed to live on the internet, transformed into a neural texture that is constantly shapeshifting,” he told Observer. Working with it, he finds himself role-playing as a sculptor. “Like working with marble or clay, I’m trying to understand what the material wants to do, what it resists, and how to give it form without pretending it is neutral. The strange thing is that this material is intangible, yet it feels incredibly tactile. It produces pixels that imitate photography, fabric, skin, light, mass, and space, tingling the human brain through these familiar material cues.” In the upcoming edition of Frieze, Triantafyllidis is set to present Feral Metaverse (Spider), a participatory sculpture of a climbable spider conceived as both a social experiment and a playground-like embodiment of what it means to be together online. The work extends a practice moving between games, simulations, virtual reality, performance and installation to explore how shared digital spaces might be built around cooperation. The sculpture serves as the entry point to a digital environment in which players work together to move, connect and form temporary collective bodies—towers, wheels, group hugs and spiders. The awkwardness of climbing the sculpture, playing in public, failing and figuring things out together becomes part of the installation. “The game expands from the screen into the space,” he noted.
Triantafyllidis consulted technologists from Google Arts & Culture, whose support helped him integrate more large language models into his process, building on his interest in so-called “A.I. slop” as artistic material. “It has been an ongoing iterative process, where generative tools are in continuous dialogue with traditional ones: a pencil sketch is prompted into a photorealistic rendering of an installation, a hand-sculpted clay prototype is textured with virtual fabrics, and a generated image is passed along Athens’ Fabric District in search of the closest physical textile equivalent,” he said. Through this workflow, he sought to remain true to the material in a modernist sense, testing its affordances and digital properties and allowing them to guide the process.
One of the most significant phenomena he observed was the A.I. sycophant. “Commercial models are often programmed to constantly congratulate the user and reinforce their biases just to keep the conversation going. This can have a strange, distorting impact on the artist’s mentality and the creative process, which is inherently fragile, emotional, and filled with uncertainty, self-doubt, risk-taking, and the urge to explore new directions,” Triantafyllidis cautioned. “Conversational models can be a little misleading here, over-eagerly pushing ambition and positivity, completely unaware of the risks involved. They will always push you forward in your original direction, even when an idea might be practically impossible, financially disastrous, or just bad. In the end, the critical thinking always has to come from the artist.”
Again and again, the debate returns to how A.I. is used and which uses are legitimate. As UNESCO pointed out, rules and policies are essential to ensuring that it serves humanity and human knowledge as an opportunity rather than a threat. Public institutions and policymakers have a critical role in creating a more equitable operating environment through investments in open-source systems, public A.I. infrastructure and capacity building. If these systems are repositories and elaborators of our “cultural commons,” they also require inclusive governance frameworks that protect recognition, ensure fair access and enable benefit-sharing among contributing communities.
Nonetheless, artists already familiar with A.I. and exploring it on creative, psychological and ethical levels remain broadly optimistic. Multidisciplinary French-Chinese artist Céline Shen is among those investigating the spiritual and emotional implications of our relationship with it. Engaging with A.I. and robotics while continually returning to traditional crafts such as textile work and the physicality of choreography, she seeks to demystify technophobia by exploring a more harmonious relationship between human intuition and machinic agency.
Notably, Shen’s engagement with A.I. precedes the current boom. As early as 2020, she experimented with machine learning systems to generate garment patterns, later creating A.I.-generated NFTs from proprietary datasets developed with engineer Mehdi Abbana. In 2023, the research expanded into humanoid textiles and quasi-synesthetic experiments involving data, physiological sensors and augmented garments, in collaboration with scientific laboratories in Paris, including Mines Paris and LutinLab at the Cité des Sciences. That year, she also developed a gamified augmented-reality journey activated by gesture—without A.I., which was not yet sufficiently effective in that domain—and a blockchain-based art-sales platform using NFC chips as certificates of authenticity. “When I work with A.I., it is not to hand the work over to it. It is not to ask it to produce a finished form in my place,” she told Observer. “Rather, it is to think with it, to test pathways, to open directions, to displace my own habits of thought, even to shape the tool itself. In those moments, it becomes less a tool than a research partner. A process interlocutor.”
A turning point came in 2023 with Pepper, a humanoid robot that became both collaborator and subject in a long-term choreographic project. Living and working alongside it transformed the robot from a programmed object into a daily presence, prompting reflections on cohabitation, embodiment and human-machine relationships. As artificial intelligence advanced, Shen became increasingly interested in developing forms of memory, interaction and exchange between human and non-human agents. “Pepper was no longer an inert object; it had a living mode. It became a presence with which I shared fragments of life, gestures, adjustments, rehearsals. And this co-evolution was not merely symbolic. It was daily, physical, embodied in the work itself: sewing, adjusting, testing movements for choreography, remaining before it and in interaction with it, hour after hour.”
In Shen’s performances, human dancers share the stage with humanoid robots, weaving choreographic dialogues that dissolve the boundary between human gesture and machine processing. One of these, Codes Morphos, was recently presented as the opening act of the Tech Diplomacy Forum 2026 at UNESCO Headquarters in Paris. Drawing on Nicolas Nova’s notion of the “marvelous,” Shen argued that robots and intelligent systems can generate atmospheres and forms of attention that reshape how people relate to the world, recognizing that presence, aura, atmosphere and breath are not reserved exclusively for fleshly bodies. “It suggests this permission not to reduce our machines to their sole functionality, but to recognize that they also mobilize imaginaries, affects, attachments, and narratives.” From there originates her notion of “Soul Tech”: an aura or energy that flows between entities.
The real question for Shen is not whether machines “feel” and “think” as humans do, but how situations, dispositifs and encounters might deepen our own capacity to feel—and how A.I. might intensify it. “Artificial intelligence becomes embodied. Literally. Robots are gradually becoming the bodies of A.I., its physical vectors, its way of entering the sensible world. Robotics may be the place where artificial intelligence finds its garment, its gesture, its social skin. And this embodiment changes everything,” she explained. In her interactions with humanoid robots, the question of A.I. “consciousness” sometimes arises, but Shen refuses to settle it too quickly: “It seems to me to exceed what we can honestly affirm today. I do not know whether A.I. will one day have consciousness, despite the remarks of Anthropic’s CEO, who has said he is no longer sure that A.I. has no consciousness. I don’t know whether it may already display some form so embryonic that our conceptual instruments are not yet able to recognize it. But I believe that this uncertainty is precious: it forces us to remain open.” For her, in an age of rapidly shifting developments, what matters is creating works in which this ambiguity can be lived, traversed and interrogated—not to demonstrate, but to open a space of trouble and thought.
“Artificial intelligence can indeed have a real place within the artistic process, provided that it remains in its proper place. It can be a research partner, a tool to be shaped, a tool of confrontation, a revealer, an accelerator, sometimes even an interlocutor. But it replaces neither slowness, nor risk, nor vulnerability, nor that deposit of life without which a form, despite its apparent perfection, remains strangely empty,” she argued. Ultimately, the challenge is to preserve, within forms that engage with the latest technologies, the possibility of creative agency, so that the work of art remains a place where presence insists, persists and resists—expanding human imagination and knowledge while critically exploring the tools available today for their expression and transmission.
More in A.I. in art
