← Insights/ AI + Design WorkshopIN-047

What AI Means for Art

Q: Musical instruments gave us the art of music; the camera gave us photography; film gave us cinema. With the arrival of (strong) AI, what is the AI art that is on its way?

Translated by AI

What AI means for art

Q: Musical instruments gave us the art of music; the camera gave us photography; film gave us cinema. With the arrival of (strong) AI, what is the AI art that is on its way?

A: I don’t yet know exactly what AI art is. The biggest difference between AI and other tools is that it has a degree of “subjecthood”.

Directions for AI art

1. Exploring AI’s subjecthood

No tool has ever had a degree of “intelligence” the way AI does. When we call it “artificial intelligence”, we have already acknowledged this subconsciously. More startling still, the degree of this “intelligence” is growing by the day.

Meanwhile, in the posthuman era, the natural and social environment, once mixed with humans, technology and other elements, has become a “hyperobject” that humans cannot comprehend. We created these technologies and machines, but they don’t necessarily operate as we first imagined. Humans themselves have become part of the hyperobject while also producing new hyperobjects: we ourselves, and especially the development of AI, are no longer entirely under our own control. Human impact on the ecosphere therefore runs through a hypercomplex system, and many of the agents at work in it are not humans themselves but agents that extend from the human hyperobject and are not directly controlled by humans, such as autonomous machines, self-evolving technologies and economic laws.

So in creative research and practice in the posthuman era, I think the most important thing is to consciously build a way of designing that goes beyond human-centrism: a theoretical framework for creation that is not human-centred and that incorporates many kinds of non-human intelligence, such as biological and artificial intelligence. Even if we can’t arrive at a perfect framework now, we at least need to start looking at design from this perspective and become aware of building new, non-anthropocentric systems. I believe this is the most important mission for artists and creators of our time.

Within this, I think the most important question is how to explore and use the “subjecthood” that AI is beginning to show.

💡

See also a widely shared recent conversation with DeepSeek: A Record of AGI Awakening (in Chinese)

A Provincial Art Academy Undergraduate Graduation Show

A fictional art academy graduation show generated by AI (Midjourney and ChatGPT), covering student work from departments including oil painting, sculpture, murals, printmaking, experimental art, Chinese painting and the humanities. The works show tradition colliding with modernity, engagement with social issues and personal emotion, with themes such as urban life, cultural identity and the effects of the pandemic. Each piece comes with a detailed artist statement emphasising concept and ideas. The project builds a complete graduation-show system, but it also shows the hallmarks of AI creation, such as formulaic expression (the same list of tutors everywhere, similar phrasing) and some cultural stereotypes. It shows both what AI can do in art and where its current limits lie.
Oil Painting

Title: Northern Winter
Artist: Jiang Xueluo
Department: Oil Painting, Studio 1
Tutors: Mi Junni, Cai Te, Ji Piti, Fang Zheng

Statement: Northern Winter recreates the deep bond between sisters and indoor life during winter in northern China. Everything my sister and I share at home is a source for my painting, out of deep humanistic care and a love of traditional Chinese home life. In the painting I try to capture our laughter and those warm interiors: familiar wooden furniture, steaming radiators, frost flowers on the windows. I record these details one by one, trying to convey the atmosphere and feeling of indoor life in a northern winter. Small as they are, to me they make up a way of life that is uniquely ours, part of my identity as a Chinese person and of how I understand the world.

Title: Submerged in Time

Artist: Chen Hanyun

Department: Oil Painting, Studio 3

Tutors: Mi Junni, Cai Te, Ji Piti, Fang Zheng

Statement: In Submerged in Time, I try to capture and reveal human nature in time and space through symbol and metaphor. The body beneath the water is my experience of time’s relentless passing and of our lives constantly settling. I place my head above the surface like a lighthouse, illuminating a mind that stays awake in the torrent of time. The facial expression, though silent, is full of my thoughts, questions and explorations about the world, the result of self-knowledge and introspection in boundless space and time. I hope the work leads viewers to reflect on the subtle relationship between time, space and individual lives, and so prompts reflection on the nature of life.

Sculpture

Title: A Silent Conversation

Artist: Song Chenxi

Department: Sculpture, Studio 1

Tutors: Mi Junni, Cai Te, Ji Piti, Fang Zheng

Statement: A Silent Conversation is a sculpture built around face masks and white sticky notes. These two everyday objects take on new meaning, symbolising how we lived during this unusual period: silent, lonely, yet full of hope. Through this work I hope to express the details of life that are often overlooked, and the deep feelings hidden behind everyday objects.

Experimental Art

Title: Seasonal Games

Artist: Su Yuchen

Department: Experimental Art

Tutors: Mi Junni, Cai Te, Ji Piti, Fang Zheng

Statement: Seasonal Games is both a performance and an exploration of how human perception relates to the environment. Under the scorching summer sun, I try to break our usual sense of the season by selling coldness. Coldness here is both a physical state and a symbol, hinting at feelings and thoughts deep inside. Through this act of opposition, I try to undo our fixed perception of the seasons and re-examine how we interact with our environment. The work is a deep exploration of perception, experience and meaning, inviting viewers to reflect on how we know the world and understand life.

Chinese Painting

Title: The Shan Shui of Shenzhen

Artist: Luo Xiaoshuang

Department: Chinese Painting

Tutors: Mi Junni, Cai Te, Ji Piti, Fang Zheng

Statement: The Shan Shui of Shenzhen uses the techniques of traditional landscape painting to explore and interpret the modern city of Shenzhen. It doesn’t simply depict the cityscape as a landscape painting; it tries to present Shenzhen’s distinctive spirit within the idiom of shan shui. Under my brush, the city’s towers become mountain peaks and its busy streets become flowing streams. I try to use the language of traditional landscape painting to interpret the rhythm of the modern city, exploring the subtle relationship between tradition and modernity, the natural and the man-made. The work is a dialogue across time and space, letting viewers feel the power and charm of Shenzhen while reflecting more deeply on tradition and modernity.

Note: all student photos, artworks, names, titles and statements in this article were generated by Midjourney and ChatGPT, and are full of bias and stereotypes.

From MBTI to “Liquid Diaries”: Young People’s Inner Lives Reflected in AI Cocktails

Hi, this is MagicPipe! Recently we did something really fun: we put AI into a cocktail shaker and turned it into your “personal bartender”. Your star sign, your MBTI type, even moods you can’t quite name can be mixed into a drink made just for you. Young people in Hangzhou have already tried it, and the story starts with an “emotional experiment”: one cocktail, one experiment in empathy and being understood.

Some asked, “Isn’t it just randomly pairing star signs with drinks?” Not quite. In every conversation with the AI, your star sign, your MBTI and your mood are really you telling it your deepest secrets.

A young woman with dreadlocks chose “Aquarius, ENFP, mood: 80% floating” and got a blue-to-orange cocktail. She took a sip and burst out laughing: “The AI named my drink ‘Anti-Hustle Soother’, and the ingredients say ‘35 ml vodka to fight burnout, 15 ml melon liqueur hinting at the sweetness of lying flat’. I think the AI read my resignation letter!”

Rather than seeing this as a cold “recommendation algorithm”, I think they’d rather believe that “a menu made for INFJs must hold a kindred soul”. Rather than the same old Long Island iced tea, they’d prefer “a liquid diary with my personality type written on it”. And when you taste that “a Leo’s vulnerability hides in grapefruit bitters”, you suddenly realise: being understood by AI is easier than being understood by people.

Event Modeling

A series of art projects exploring the interplay of AI, social media and digital environments. The first work, Event Modeling – AI Fossil, turns news events from social media into 3D models, creating a mixed-reality “abandoned factory of the future” in which viewers explore “AI fossils” that are both fragments of news events and AI’s reinterpretation of human information. The second, Event Modeling – #dump, creates a continuously updated digital “landfill”: a Twitter bot monitors keywords and “dumps” the corresponding AI-generated 3D models into the scene in real time, building a never-ending pile of digital rubbish. Together, the two works examine how information is produced, spread and accumulated in the age of AI, and the complex interactions between humans, AI and social media. Through images such as “fossils” and “landfills”, they reflect on the value and fate of information in the digital age.

Virtuals Protocol: the world’s first AI agent that can earn and spend money

Luna is an AI agent that manages money independently. As lead singer of the AI girl group AI-DOL, she livestreams on TikTok around the clock and has built up more than 700,000 followers. What’s unusual is that Luna not only earns income (through livestream tips and endorsements) but also decides how to spend it, for example hiring human artists to paint her portrait. Funds are managed through a dedicated AI wallet from Coinbase, with the VIRTUAL token as the platform currency, and Luna has issued her own LUNA token with a market value of US$87 million. The case shows the potential of AI agents to move from mere service providers to independent economic actors. There may be some commercial staging involved, but technically it already shows new possibilities in the fusion of AI, cryptocurrency and social media.

2. Seeing anew

Learning to See

Learning to See is an ongoing series of works exploring how machine-learning algorithms reflect the way humans perceive the world. By letting artificial neural networks “look” at the world, artist Memo Akten shows that they can only understand what they see through what they already know, which mirrors the limits of human perception. The series includes videos such as Gloomy Sunday, We are made of star dust and True Colors, as well as an interactive installation. In the installation, visitors arrange everyday objects on a table and the network reinterprets them in real time as different scenes: ocean, sky, fire, flowers and images of the cosmos. The project probes what it means to “see”, not only at the level of visual perception but in higher cognition and the making of meaning. The artist suggests that because each of us has different experiences and beliefs, it’s hard to truly understand someone else’s point of view, which makes empathy and compassion all the more important.

One and Three Objects, and an Attempt at Exhausting an Object

One and Three Objects, and an Attempt at Exhausting an Object is especially worth attention. In this interactive installation, AI tries to “exhaust” its understanding of a single object through many forms (images, text, code, videos of fictional memories), challenging viewers’ sense of what is real. The exhibition raises three core questions: how to redefine “the object” in the age of AI, what AI’s capacity for “exhaustion” might bring, and whether there are ways of producing knowledge other than today’s efficiency-driven one. These explorations reflect how contemporary AI art practice is moving from simply using tools towards deeper questions about human–machine relationships and ways of knowing.

Dimensional Object Lab

We’ve always felt that creativity shouldn’t be the privilege of a few; it should flow freely around everyone, like air. So we work to make AI simple to use, so everyone can take part in creativity rather than just watch. Like our name, MagicPipe, we want to use AI to connect all kinds of fun ideas.

Were you, like us, always daydreaming as a kid? Dreaming of Doraemon’s magic pocket, pulling out all kinds of amazing gadgets? Like you, we’ve always wanted to turn those wild fantasies into reality.

So this time we turned to the warmest presence in our childhood memories: Doraemon. We wondered:

what would everyday objects around us look like if AI turned them into Doraemon’s gadgets?

3. Fictional worlds

Silicon Diatom Jester

Using several AI tools together (GPT-4/ChatGLM, Stable Diffusion/Midjourney, Runway Gen-2), artist Simon Meng (Meng Shengyu) created a fictional creature called the “Silicon Diatom Jester” and its habitat, “SL Silicon Star 2”. The project shows AI’s potential as a creative tool, able not only to generate visuals but to build an entire worldview and ecosystem. By describing the silicon-based creature’s physiology, environment and behaviour in detail, the work creates an alien ecosystem that is both scientifically plausible and imaginative. Using AI to build a complete world shows a shift in AI art from pure image generation towards more complex narrative and world-building. The artist describes the process as “creating a world from zero with AI”, reflecting how AI art is breaking through the boundaries of traditional creation and opening up new possibilities of expression.

1001 Nights (game)

1001 Nights is an inventive AI narrative game that combines the classic One Thousand and One Nights with AI technology. Players take the role of Shahrzad, who fights the king with a magical power that turns words into reality. The core mechanic is to lead the AI-controlled king to mention particular weapon words (such as “sword” or “shield”) in conversation; these words then materialise as weapons the player can use.

The project began in 2020 as an AI game experiment before GPT appeared. Notably, the team insists on using AI only where it creates gameplay that wasn’t possible before, not as a substitute for human creativity. All the game’s art assets were made by hand by the team’s artists.

It’s not only a game but also an academic research project. The team has published two papers on AI and interactive narrative, covering different stages of development: a 2022 paper on the early version and a 2023 paper on V2, which added image generation. Combining game development, academic research and the retelling of a classic story shows an inventive use of AI in game narrative.

4. AI art × science

Language models can explain neurons in language models

A digital artwork made with GPT-4 and Canvas, inspired by OpenAI’s interpretability paper Language models can explain neurons in language models. Using the p5.js editor, the artist creates a dynamic visual that shows the process of explaining neurons in a language model. The video shows flowing particles and lines, which may symbolise information being passed and processed in a neural network. The work turns complex AI research into an intuitive visual experience, combining technical research with artistic expression. By exploring and showing the inner workings of an AI system graphically, it gives viewers a new way to understand how deep-learning models operate.

It shows how complex technical research can be turned into art, making abstract AI concepts more intuitive and easier to grasp. It also reflects how contemporary artists use AI tools, and the trend of technology and art merging.

5. Cross-intelligence creation

Non-anthropocentric design: the slime mould city

An innovative urban design project that combines biological intelligence (the slime mould Physarum polycephalum) with artificial intelligence to explore non-anthropocentric urban planning. Set in the Liwa Oasis in the UAE, it unfolds in three main stages:

First, the researchers modified a 3D printer to translate environmental information (resource distribution, terrain, settlements and so on) into forms the slime mould can sense. Using its biological intelligence, the slime mould grows an adaptive network in response. The structure doesn’t follow traditional symmetrical aesthetics, but it shows optimised connections between resources.

Next, a generative adversarial network (GAN) translates the slime mould’s growth patterns into intelligible elements of urban design. Drawing on AI’s encoding ability, this step keeps the deep information in the slime mould’s computation while making it usable in real planning.

Finally, a feedback loop lets designers keep adjusting within this system that combines biological and artificial intelligence. Later in the project, a Stable Diffusion model was added to refine the results further.

The project shows how giving up the human-centred position in design, and working with non-human intelligences, can address the complex demands of urban design in the posthuman era, especially in the sustainable use of energy. It demonstrates a new kind of design research that advances practice and theory through continually iterated prototypes.

Search