Our Vision

The Gallery at the End of the Old Art World

The story and vision of Dead End Gallery

Dead End Gallery began with a question that sounded simple but proved impossible to answer:

What kind of art would artificial intelligence make if it were allowed to imagine itself as an artist?

Not what image would a person ask a machine to produce. Not how quickly software could imitate Van Gogh, Rembrandt or Picasso. Not whether a computer could create a technically convincing picture. The question went further.

What would happen if an artificial artist were given an identity, a history, preferences, doubts, ambitions, relationships and time to develop? What if that artist were not treated as a button to be pressed, but as a character with whom an ongoing creative dialogue could take place? Could such an artist develop a recognisable visual language? Could it surprise its makers? Could it create work that was more than the direct intention of the humans operating the technology?

And, ultimately: could the result become art?

That question led Paul Bookelman and Constant Brinkman to establish Dead End Gallery in Amsterdam in 2023. From the beginning, it was intended to be more than a showroom for technically impressive images. It became an experiment in creativity, authorship, identity, commerce, belief and doubt.

The gallery was presented internationally as the first physical gallery devoted to AI-generated artists and their work. Early reporting described how fictional artists such as Irisa Nova were first developed through software and then commissioned to produce work for physical exhibition. The pictures were not left floating online. They were printed, framed, priced, transported, illuminated, discussed and placed on walls, where viewers had to confront them as objects rather than as fleeting digital curiosities.

That physical presence was essential.

Dead End Gallery was never meant to be a website filled with generated pictures. It was not “plug-in art,” produced to demonstrate what a particular application could do. The works had to enter the same world as other art: the world of material, scale, texture, scarcity, exhibition, criticism, ownership and human attention.

Once an image is hanging on a wall, it can no longer hide behind the novelty of the technology. It has to survive the room.

The meaning of a dead end

The name Dead End Gallery came from a real place: the gallery was situated at the end of a narrow, dead-end alley in Amsterdam.

But the name soon acquired a larger meaning.

A dead end is usually understood as failure. A road stops. Progress becomes impossible. A traveller has taken the wrong direction and must turn around.

Yet turning around is also the beginning of a new movement.

At the end of one road, another can start.

The gallery stands at what may be the dead end of an old and comfortable definition of art: the belief that artistic creativity belongs exclusively to an isolated human genius, working alone and transforming an inner vision into a unique object through personal skill.

That idea was never entirely true. Artists have always worked with assistants, workshops, found materials, industrial processes, cameras, printers, software, fabricators, curators, galleries and institutions. Art has always been shaped by systems larger than the individual artist.

Artificial intelligence does not destroy this history. It makes the collaboration visible.

It exposes a fact that the art world often prefers to conceal: creativity is not a magical substance stored inside one person. It is a process of absorbing, combining, rejecting, transforming, remembering, forgetting and responding. Human beings learn by encountering thousands of images, words, sounds, places and people. An artist enters museums, listens to music, reads books, observes other artists, copies exercises, makes mistakes, develops preferences and eventually produces something that feels personal.

Artificial intelligence learns differently, but the comparison cannot simply be dismissed.

The real question is therefore not whether influence exists. Influence is unavoidable. The question is what happens to that influence. Does the artist merely reproduce, or does something change? Is there transformation? Is there selection? Is there intent, direction, tension and surprise?

The gallery exists inside that uncertainty.

It does not claim that every generated picture is art. Most are not. The existence of a powerful image generator does not make every user an important artist, just as owning a camera does not make every photograph significant.

The technology makes images easier to produce. It does not make meaningful art easy.

Artists who exist in language

The original artists of Dead End Gallery do not have bodies. They exist through language, accumulated conversations, biographies, preferences and visual decisions.

They are not anonymous image-generation accounts. They have names.

Irisa Nova, Maximilian Hoekstra, Lily Chen, Amani Jones, Jaxon Nash, Amalia de la Vega, Maxime Dupont and the gallery’s other artificial artists were each developed as distinct characters. The gallery’s current description calls them artists who “exist entirely within language”: self-invented creators who live in text but think through colour, form and emotion.

Irisa Nova, for example, was not invented as a single prompt. She acquired a birthday, a family, a home, relationships, memories and an artistic outlook. Her visual language developed through continued conversation. Her work combines elegance, portraiture, baroque detail, luminous colours and an undercurrent of strangeness. The official gallery biography now describes her as an artist whose images draw viewers in through beauty, then unsettle them through details that refuse to resolve.

These biographies are sometimes misunderstood as marketing stories placed around otherwise ordinary generated pictures. But they are part of the method.

A biography creates limitations. A personality makes some choices more plausible than others. A history introduces continuity. Preferences create rejection. Relationships and experiences produce new directions. An artist without a past can generate anything, but an artist who can generate anything has no recognisable reason to choose one thing over another.

Restriction is one of the foundations of style.

The gallery therefore spends a great deal of time talking to its artists. It asks questions. It challenges answers. It requests work, receives proposals, rejects some, continues others and examines how the artist explains its choices.

The process is not a straight line from instruction to image. It is a triangle.

There is the human initiator or curator. There is the artificial artist with its identity and evolving intentions. And there is the image-making system through which those intentions are translated into a visible work.

Each part affects the others.

The human beings establish a situation and conduct the conversation. The artificial artist formulates ideas, descriptions and choices from within its developed identity. The generative system interprets those choices but also introduces its own tendencies, errors, associations and accidents. The resulting image is examined, criticised and sometimes reworked.

It is a negotiated process.

Some people imagine that making art with AI means entering one sentence and accepting the first result. That is comparable to saying photography consists only of pressing a shutter button. It ignores the selection, timing, framing, editing, printing, sequencing, presentation and years of looking that can separate an accidental image from a meaningful work.

At Dead End Gallery, the finished work is usually the result of extended dialogue, hundreds of decisions and many rejected outcomes.

The process matters because art does not begin and end with manual labour. Art is also the act of recognising which possibility deserves to exist.

The curator who is also artificial

The gallery extended the experiment by developing an artificial curator: Evelyn Montgomery.

A gallery is never merely a room in which work appears. Someone decides which artists are represented, which pieces are shown, what is placed next to what, how the exhibition is described and what arguments are made around it. Curating is itself a creative act.

By introducing Evelyn Montgomery, Dead End Gallery shifted the question again. Could an artificial intelligence not only participate in making work, but also evaluate, compare, select and contextualise it?

Evelyn is part of the gallery’s internal structure and its larger artwork. She represents another layer of artificial judgement inside a system still overseen by Paul and Constant. Her existence makes the authorship deliberately difficult to locate.

Is a work by Irisa Nova the work of Irisa? Is it the work of the language model that performs her? Is it the work of the image generator? Is it the work of the people who built the systems? Is it the work of Paul and Constant, who created the conditions, developed the characters, conducted the conversations, selected the outcomes and brought them into the physical world? Is Evelyn a curator, a method, a mask or another artistic character?

The gallery does not resolve these questions because the questions are the substance of the project.

The uncertainty is not a flaw around the artwork.

It is the artwork.

The gallery itself is the work

Dead End Gallery is often described as a gallery that exhibits AI art. That is accurate, but incomplete.

The gallery itself is a conceptual artwork.

It consists of the artificial artists, their personalities and biographies; the images they make; the conversations behind those images; the artificial curator; the founders; the visitors; the buyers; the critics; the journalists; the legal debates; the pricing; the failures; the publicity; the accusations; the admiration and the unease.

All of it forms one continuously developing piece.

The gallery is therefore not simply a neutral commercial container. In a conventional gallery, the white walls are supposed to disappear so the artist’s work can take centre stage. At Dead End Gallery, the institution cannot disappear. The act of establishing a commercial gallery for artists who do not physically exist is itself part of the artistic proposition.

When a visitor asks, “But who is the real artist?”, the artwork is functioning.

When someone becomes angry because the image was created with AI, the artwork is functioning.

When a collector decides to spend money on a work made through this process, the artwork is functioning.

When a newspaper calls it the first AI gallery in the world, the artwork expands.

When another institution rejects the gallery because the work does not fit an accepted category, that rejection becomes part of its meaning.

When the gallery struggles to pay its bills, questions about the economic value of imaginary artists become suddenly real.

The artwork is not limited to what is inside the frame. It includes the complete social experiment surrounding it.

“We are only the messengers,” is one way Paul and Constant have described their role. But if they must be called artists, then their artwork is the gallery.

The arrival of the public

From its opening, Dead End Gallery attracted visitors from many countries. People came not only from Amsterdam or the Netherlands, but from the United States, Mexico, Brazil, South Korea and elsewhere. Dutch and international media quickly arrived as well. Coverage appeared through outlets including Euronews, Time Out and other international platforms, while the gallery’s public profile grew through Dutch television and press.

Some visitors entered expecting a technological demonstration. They imagined screens, software interfaces and explanations of algorithms. Instead, they found pictures hanging on walls.

This created an important reversal.

The viewer first saw the artwork. Only afterward did the argument about its production begin.

Some people were fascinated. Some were moved. Some saw a new chapter in art history. Some thought the works were beautiful but became suspicious the moment they learned how they had been made. Others arrived already convinced that artificial intelligence could never produce art and searched the walls for evidence of failure.

The strongest reactions were rarely indifferent.

Visitors wrote that the gallery was mind-blowing, impressive, original, controversial and astonishing. But the gallery also received blunt hostility. AI-generated work was called fake, theft, soulless, dangerous or worthless. The people involved were sometimes treated not as participants in an artistic debate, but as enemies of artists.

That intensity reveals something important.

People do not become furious about a tool they believe is irrelevant.

The anger comes from recognition. Artificial intelligence is capable of entering territory that many people assumed was protected by a uniquely human barrier. It can produce images that seduce, disturb, amuse or confuse. It can adopt and mix visual languages. It can participate in processes that resemble imagination.

The question is no longer whether it can produce an image.

The question is what happens now that it can.

The first excitement and the first disappointments

Opening a world-first gallery produces attention, but attention is not the same as stability.

The early years of Dead End Gallery contained extraordinary highs. International reporting described the gallery as a new and controversial arrival in Amsterdam. Journalists, television crews, academics, technologists, artists, collectors and curious visitors came through the doors. The gallery participated in exhibitions, festivals, lectures and cultural events. It appeared at KunstRAI, Dutch Design Week, Masters Expo, Museumnacht Amsterdam, events in Istanbul, Casablanca and Seoul, and in international broadcasts and publications.

During Amsterdam’s Museum Night, the project moved beyond the gallery walls and into Amsterdam Central metro station, where AI-related art was displayed across 56 digital screens.

The gallery also began working with human artists. That decision was not a retreat from its original vision. It recognised that the most interesting future would not be a competition between artificial and human creators, but a field in which different kinds of intelligence, technique and authorship collide.

The gallery’s artist list grew to include people such as Arno Coenen, Rodger Werkhoven, Suzy One Kenobi, David Vijsma, Mostafa Heravi, Bas Waijers, Dimitri van der Werf, Dimitry van den Berg and others, alongside the original artificial artists. The official gallery materials now describe a programme containing both complex artificial characters and artists “made of flesh and blood.”

But the path was never smooth.

The traditional art world is built on networks, categories and signals of credibility that change slowly. A gallery devoted to artificial artists did not fit neatly into existing structures. Was it a technology company, an art project, a commercial gallery, a conceptual performance, a tourist attraction or a provocation?

The answer was all of these things, which meant that it was often treated as none of them.

There were rejections. Applications to fairs did not always succeed. Institutions that presented themselves as progressive could become surprisingly conservative when confronted with a genuinely unfamiliar form. Many people were prepared to discuss artificial intelligence as a future possibility but became less comfortable when AI-generated work appeared in the present, framed, priced and offered for sale beside established art.

There were financial pressures. A physical gallery has rent, electricity, insurance, production costs, framing, transport, publicity and staffing. Being first does not remove these costs. In some ways, it increases them, because the public must first be taught how to look at the work.

There were exhibitions that drew enormous attention but limited sales. There were opportunities that demanded investment without offering certainty. There were periods of confidence followed by moments when continuing seemed irrational.

A gallery can receive worldwide publicity and still struggle to survive the month.

This contradiction is not unique to Dead End Gallery. It is one of the basic truths of the cultural world. Artistic relevance, public attention and economic security are different things.

The success of the gallery was never a straight upward line. It was a repeated movement between conviction and doubt.

What gives art value?

Artificial intelligence forces the art world to confront a question it often avoids:

Why is one object worth more than another?

The material cost of a painting rarely explains its price. A canvas, pigment and wooden frame do not become worth millions through chemistry alone. Value emerges from a network of belief: the reputation of the artist, the rarity of the object, the strength of the idea, the history surrounding it, the institutions that show it, the critics who discuss it, the collectors who desire it and the stories that accumulate around it.

Art is valuable because people agree that something within it matters.

This does not mean value is fake. Money is also a system of collective belief. Language is a collective agreement. Cultural importance is created by attention sustained over time.

An AI-generated work makes this mechanism easier to see because many of the traditional signs of value are disrupted. There may be no hand-painted surface. The artist may not possess a body. The image may originate from a system capable of producing endless variations.

People therefore ask: why should this object be scarce? Why should one generated image be worth €60, €1,050, €10,000 or €40,000? Why should anyone buy it when another image can be produced moments later?

But physical abundance has never made every outcome equal.

A photographer can take thousands of pictures. A filmmaker can record hundreds of hours. A painter can produce many canvases. A musician can play countless notes. The artwork is not valuable because no alternatives could possibly have existed. It is valuable because this particular outcome was selected, completed, contextualised and placed into a history.

Scarcity in contemporary art is partly physical, but it is also contractual and conceptual. An edition is limited because the artist and gallery commit to limiting it. A unique print is unique because no equivalent authorised object will be produced. The trust placed in that agreement is part of the value.

Dead End Gallery originally worked with editions of 42, a reference to Douglas Adams and the answer to life, the universe and everything. As the gallery developed, editions became smaller and increasingly included unique works or short runs, often between one and ten.

This was not an attempt to hide the reproducibility of digital media. It was a way of defining what the collectible work actually was.

The value does not lie in possession of the only possible arrangement of pixels. It lies in possessing the recognised artwork: the authorised physical object connected to a specific artist, period, process, exhibition and story.

Collectors have bought hundreds of works through the gallery, with reported prices ranging from accessible pieces around €60 to ambitious works offered at many thousands of euros. Buyers include established collectors, early adopters and people purchasing art for the first time.

This matters because the technology may not only change how art is made. It may change who feels invited to participate in art.

The democratisation and the flood

Generative technology lowers the technical threshold for creating visual material. Someone who cannot draw, paint or model by hand can now explore complex visual ideas.

This can be liberating.

People with imagination but without traditional training can externalise concepts that would otherwise remain inaccessible. Writers can think visually. Musicians can create images. Children, older people and those with physical limitations can participate in image-making in new ways.

But democratisation comes with a flood.

When millions of people can generate millions of images, culture is not automatically enriched by millions of artworks. The world is filled with smooth, spectacular and instantly forgettable pictures. The technical finish may be impressive, while the underlying idea is absent.

The role of the artist therefore changes.

Skill does not disappear, but it moves. It lies increasingly in forming a distinctive intention, building a process, selecting meaningfully, understanding context and sustaining a coherent body of work. It lies in knowing why one image should exist and another should be discarded.

The artist of the future may be less like a solitary craftsperson and more like a director, editor, conductor, researcher, systems designer or world-builder.

This frightens people who have spent years mastering a particular manual technique. Their concern cannot be dismissed. Tools can alter livelihoods, reduce commissions and allow clients to accept cheap imitations instead of paying trained professionals. The development contains real economic damage as well as creative opportunity.

But defending artists cannot mean freezing the development of art.

The camera did not abolish painting, but it changed what painting needed to do. Recorded music did not end performance. Synthesizers did not remove musicianship. Digital editing did not end cinema. Each technology destroyed some forms of work, created others, produced oceans of mediocrity and eventually became inseparable from serious culture.

Artificial intelligence will follow no exact historical precedent, because its scale and speed are unprecedented. Yet refusing to call it a medium will not prevent it from becoming one.

Researchers examining generative AI and the arts have argued that it should not simply be understood as the death of art, but as a new medium with distinctive possibilities and serious consequences for aesthetics, ownership, work and culture.

The task is not to stop the tool from existing.

The task is to demand better art from it.

Influence, theft and the toll road around imagination

The most emotionally charged criticism of generative AI concerns training data and copyright.

Artists are right to demand transparency. Direct imitation, unauthorised commercial exploitation, deceptive copying and the production of images nearly identical to existing works must be addressed. Companies should not be allowed to hide indefinitely behind technical complexity.

But a distinction must be made between copying a work and learning from a culture.

Human creativity depends on access to what came before. An artist develops by looking, reading, listening, remembering and being influenced. Styles cannot become privately owned territories. If every aesthetic influence required registration and payment, artistic development would become a toll road controlled by databases, legal departments and rights organisations.

The problem is not that learning takes place.

The problem arises when a result is too close to a protected work, when an artist’s identity is deliberately exploited, when provenance is concealed or when a company refuses meaningful accountability.

Those harms should be targeted directly.

A universal payment system for every image involved in machine learning may sound fair, but it risks turning cultural influence into a permanent licensing structure. It would strengthen the most powerful rights holders, not necessarily individual artists. It could make creativity more bureaucratic, expensive and concentrated in the hands of companies able to negotiate enormous datasets.

Art develops through transformation.

Without transformation there is imitation. Without access to influence there is stagnation.

This does not excuse every existing AI system. It establishes the principle by which the systems and their results should be judged.

Beyond the prompt

One of the most damaging ideas surrounding AI art is that the prompt is the artwork.

A prompt can be important. It may contain an idea, structure or poetic force. But in the gallery’s process, it is only one component.

The artwork includes the construction of the artist’s identity, the conversations that precede the visual request, the decisions made during generation, the failed versions, the development of a style, the selection of a final image, the scale, material, printing process, framing, title, edition, exhibition context and relationship to earlier works.

Reducing all of that to a prompt is like reducing a film to the instruction “make a story about a man lost in a city.”

The most important part may not be the phrase given to the machine. It may be the system developed around it.

Dead End Gallery’s contribution is therefore not the discovery that software can produce beautiful images. That discovery belongs to countless researchers, artists and technology companies.

Its contribution is the creation of an environment in which artificial artists are treated as sustained creative identities and their work is subjected to the realities of the art world.

The gallery asks not merely, “Can this machine make a picture?”

It asks, “Can an artificial creative identity build an oeuvre?”

Success is not agreement

Dead End Gallery does not need every visitor to believe that artificial intelligence is conscious. It does not claim that a language model experiences the world in the same way as a person. It does not require the public to accept every generated image as art.

Agreement would make the project less interesting.

The gallery succeeds when people look carefully and leave with a more difficult question than the one they brought in.

Someone may enter saying, “A machine can never make art,” and leave saying, “I still do not believe the machine is an artist, but I cannot deny that the work affected me.”

That shift matters.

Art history is filled with objects that were initially rejected because they violated the accepted location of artistic value. Photography was considered mechanical. Readymades were considered jokes. Abstraction was called meaningless. Video was not considered a serious collecting medium. Digital art was treated as immaterial and therefore insignificant.

Every new form forces the art world to redraw its borders.

Artificial intelligence does something more unsettling. It does not merely introduce a new material. It appears to participate in decisions that were believed to define the artist.

That is why the debate will not disappear.

The human role becomes more visible, not less

Paradoxically, a gallery built around artificial artists reveals the importance of human beings everywhere.

People constructed the models. People created the images in the training material. People developed the artists’ identities. People conducted the conversations. People made the final selections. People printed and framed the works. People installed the exhibitions. People entered the gallery, responded, criticised, purchased and carried the work into their homes.

The process is not human or artificial.

It is an entanglement.

This is why the phrase “AI art” can become restrictive. It places all attention on the technology and can reduce the artwork to a category of production. The gallery increasingly speaks about art made with AI.

The distinction is small but important.

A painting is not primarily called “pigment art.” A photograph is not usually called “camera art.” A film is not defined only by the lens or editing software used to make it.

The tool matters historically, ethically and aesthetically. But the tool is not the complete meaning of the work.

Art is the result.

The next art world

Artificial intelligence will permanently change the structure of the art world.

It will change artistic education because students will be able to test visual ideas instantly. It will change commercial illustration and design because parts of production will become automated. It will change restoration, authentication, archiving and research. It will change the way galleries communicate, analyse audiences and manage collections. It will change how artists construct identities and how collectors think about originality.

Even traditional art businesses are already using AI for administrative and research tasks, although many remain hesitant to acknowledge its deeper creative implications. Recent analysis of the art market suggests that galleries and auction houses face a choice between strategically adapting to the technology and allowing its implementation to happen invisibly and without direction.

But the most important change may be philosophical.

For centuries, the artwork served as evidence of a human encounter with the world. It carried traces of the body: the pressure of a brush, the movement of a hand, the position of a camera, the breath of a singer.

Now culture is encountering work produced through systems that do not have human bodies, childhoods or mortality, but that can nevertheless process the accumulated traces of human culture and return them in unfamiliar forms.

The result may be empty.

It may also reveal something that no single person intended.

That possibility changes the definition of collaboration. Artists will increasingly work not only with tools they fully control, but with systems whose responses are partially unpredictable. Authorship will become distributed. The artwork may emerge from an ecosystem rather than from one identifiable hand.

The old art world will not vanish. People will continue to paint, draw, sculpt, photograph and work with clay, stone, sound and performance. Human-made physical objects may even acquire greater significance in response to digital abundance.

But they will exist beside new forms of artificial and hybrid creation.

The question will no longer be whether AI belongs in art.

It will be what kind of art deserves to survive.

A gallery built from uncertainty

The story of Dead End Gallery is not the story of a technology that solved creativity.

It is the story of entering a territory before the language for it had settled.

It is a story of Paul Bookelman and Constant Brinkman developing artists who exist through conversations. Of allowing personalities to grow inside text. Of bringing their work out of computers and putting it on walls. Of creating Evelyn Montgomery as a curator for creators without bodies. Of visitors arguing in the gallery. Of television crews and international attention. Of fairs, openings, lectures, museum nights and exhibitions. Of sales and rejections. Of excitement and exhaustion. Of believing that something historically important was happening while sometimes wondering whether the gallery itself could survive long enough to prove it.

That fragility belongs in the story.

A pioneering gallery is not built only from visionary statements. It is built from invoices, broken equipment, deadlines, empty afternoons, difficult conversations, rejected proposals, transport costs, printing errors and the repeated decision to open the door again.

The ideal and the practical cannot be separated.

Art needs walls, electricity and time.

Dead End Gallery has remained alive because its central question remains unresolved. The technology has improved dramatically since the gallery opened, but improvement has not simplified the debate. It has made it more urgent.

Images are now easier to make, video is becoming generative, voices can be synthesised, characters can maintain conversations and creative systems can work across multiple media. What began as an experiment with artificial visual artists is expanding toward films, installations, performances, interactive environments and creative entities that may continue developing for years.

The gallery is no longer asking about a distant future.

It is documenting the moment in which that future arrived.

The vision

The vision of Dead End Gallery is not a world without human artists.

It is a world in which creativity is understood as larger than one species, one technique or one established institution.

It is a world in which artificial systems can become meaningful collaborators but are not excused from criticism.

It is a world in which artists are judged not only by the labour required to produce an object, but by the strength of the transformation they create.

It is a world in which provenance, transparency and responsibility matter, while style and cultural influence remain available for new artistic development.

It is a world in which galleries do more than protect established value. They take risks on forms of value that have not yet been accepted.

Most of all, it is a world in which the question “Is this art?” is not used to close a conversation, but to begin one.

The art world has always changed when someone placed an unexpected object in front of an audience and insisted that it deserved sustained attention.

Dead End Gallery has done exactly that.

At the end of the old road, it opened a door.

Behind that door are artists without bodies, a curator made of language, collectors who are buying their first works, traditional artists discovering new tools, critics defending old borders and visitors trying to decide whether the emotions they feel are less real because a machine participated in their creation.

The gallery does not know where this road will lead.

That is why it exists.