Why art still matters when machines can also make it
Dear Reader,
There is a specific argument I keep hearing, sometimes made confidently by people who do not make art and sometimes made anxiously by people who do. The argument runs like this. AI can now produce images that look like art. It can produce music that sounds like music. It can produce text that reads like poetry. Which means human art is obsolete, or at least optional, because the outputs can be generated at scale by machines with no specific human involvement required.
This argument is confused. It confuses the specific outputs of art with the specific practices that produce them. It confuses what art looks like from the outside with what art does for the specific people making it. And it treats the aesthetic dimensions of art as if they were the whole of it, when in fact they are the surface manifestation of something else that AI does not touch.
I want to write about why art still matters, in specific ways that AI cannot replicate, and why the practice of making art has become more important, not less, as machines have gotten better at producing artifacts that look like art. This piece opens Cluster D of this newsletter series, which is about specific practices that let human meaning survive the transitional environment I have been writing about. Art is the first of these practices, and one of the most important.
Start with what art actually is, because most public discussion of AI and art operates on a definition that is too narrow.
Art has been many things across history. In some cultures it has been primarily religious, made to honor gods or ancestors, and evaluated by whether it served those purposes rather than by aesthetic criteria alone. In some cultures it has been primarily craft, embedded in daily life and evaluated by whether it fit well into that life. In some cultures it has been primarily political, made to advance specific causes and evaluated by its effects. In some cultures it has been primarily private, made for the maker's own purposes and not intended for wider audience.
What all of these versions have in common is that art is a specific practice of making meaning. Not merely producing objects. Making meaning. The specific objects produced are the material trace of a specific process by which the maker worked out something about themselves, their world, or their relationship to both. The object is not the whole of the practice. The object is the shape the practice left behind.
This distinction matters when we consider AI. AI can produce objects that resemble the shapes that human art practice leaves behind. It cannot engage in the specific practice that produced those shapes. Which means AI-generated images may look like paintings, but they are not the residue of anyone working out anything. They are the outputs of an optimization process that had no specific stakes, no specific struggle, and no specific transformation for anyone.
This might not matter if art were only about the outputs. But art has never been only about the outputs. It has always been about what the outputs did for the people making them, and secondarily what they did for the people receiving them. Both dimensions depend on the outputs being produced through a specific human practice. Neither transfers to AI outputs, even when those outputs are visually or auditorily equivalent.
Let me be specific about what the practice of making art actually does for the person doing it.
First, it forces the maker to face specific things they might otherwise avoid. When you sit down to make something, you have to make specific choices. What is this about. What am I trying to say. What am I willing to look at long enough to represent. These choices are not typically resolved by the maker's conscious intentions. They are resolved through the making itself, in a way that surfaces material from the maker that would not have surfaced through pure introspection.
I know this because it has happened to me. When I sit down to work on a piece, whether writing or visual work, specific things come out that I did not know were in me. Specific memories. Specific ambivalences. Specific fears. Specific commitments. The specific process of trying to shape material forces me to confront what I actually think and feel about it, rather than what I believed I thought and felt before I started.
This is not a mystical process. It is a specific psychological one. The act of externalizing something in a specific medium requires making specific choices that reveal information about the maker to themselves. The information was already there. The making surfaces it. And once surfaced, it becomes available for the maker to work with in ways that would not have been possible before.
AI cannot do this for the person using it. AI can produce outputs that look like what the person might have produced. But the specific psychological work of making the outputs, and the specific self-knowledge that comes from that work, is not present in the AI-assisted case. The person who prompts an AI to produce art gets the artifact without the practice. Which means they get less of what art actually does.
Second, art forces the maker to develop specific capacities that transfer to other domains.
Sustained attention. Careful observation. Willingness to fail. Tolerance for slow progress. Ability to work with materials that resist you. Willingness to abandon what is not working. Judgment about when something is finished. Each of these capacities is developed through the specific practice of making art, and each of them is useful in domains far beyond art itself.
The person who has spent thousands of hours making things has capacities that the person who has spent those hours on other activities does not. This is not because art is uniquely virtuous. It is because the specific practice of making requires the specific capacities in the specific combination, and developing them through art is one of the specific paths to having them available for other work.
AI does not develop any of these capacities in the user. Prompting a system to produce an artifact does not require the sustained attention that making the artifact directly would require. It does not develop the specific tolerance for failure that iterating on a physical medium develops. It does not build the specific judgment that thousands of hours of practice builds.
Which means that the person who uses AI to produce art-like artifacts is not developing the specific capacities that making art traditionally developed. They are getting the artifact without the training. And the training is arguably where the value was.
Third, art creates specific artifacts that carry specific meaning for the specific communities that produced them.
The specific artifacts that survive from previous eras tell us something about those eras. The specific choices the makers made encode something about the specific concerns of the specific societies they lived in. When we look at a specific painting from a specific place and time, we see the specific way the maker was engaging with the specific reality they lived in. This is why art has been, historically, one of the primary sources of information about how people in previous eras actually thought and felt.
AI-generated artifacts do not have this specific quality. They are not the residue of anyone engaging with anything. They are the specific outputs of statistical patterns in training data, produced according to specific optimization objectives, with no specific stakes for any specific maker. Which means they cannot serve the specific function that human-made artifacts have served. They cannot be sources of information about how anyone actually experienced anything. They are aesthetic outputs without any specific experiential referent.
This is a substantial loss when it occurs at scale. If a significant portion of the visual and cultural production of a period is AI-generated, then the specific historical record of that period will contain a large amount of material that does not tell us what anyone thought or felt. The specific gap between what we can know about previous eras and what we can know about the current era will widen, and not in favor of the current era.
Let me tell you about something specific from my own practice.
I make visual art under the name FALSE GOD. Some of it is available publicly. Most of it is not. The specific process of making it involves specific decisions, specific mistakes, specific reworkings, and specific moments where I decide something is finished or abandon it entirely.
I have experimented with AI-assisted image generation. I have used it as a tool. What I have learned is that the specific artifacts it produces feel different to me than the specific artifacts I make directly. Not necessarily worse aesthetically. But different in a specific way that matters for how I relate to them. The AI outputs do not feel like mine, even when I have prompted them extensively. They feel like collaborations with something whose interests are not aligned with mine. The specific artifacts I make directly, even when they are more limited technically, feel like specific expressions of specific things I was trying to work out. The specific work is present in the artifact in a way that the AI output cannot replicate.
This is not just personal preference. It is a specific observation about the specific texture of different kinds of production. When you make something directly, the specific choices you made are visible in it, at least to you. When you prompt an AI, the specific choices are largely made by the system, not by you. Which means the artifact does not carry your specific decisions in the specific way that direct making produces.
I keep both kinds of work. I use AI when it serves specific purposes. I make directly when what I need is the specific thing that direct making produces. The two are not substitutes. They are different tools for different jobs. And the specific mistake that a lot of contemporary discussion makes is to treat them as substitutes.
There is a specific broader implication that I want to name.
Human meaning has traditionally been produced through specific practices. Making things. Building things. Growing things. Raising children. Caring for the sick. Teaching. Governing. Each of these practices produces specific forms of meaning for the people engaged in them, and each of them has been progressively automated over the last several decades. Different practices have been automated at different rates and to different degrees. All of them have been affected.
The specific consequence is that the specific sources of meaning that used to be widely available are becoming less available. Not because they are being taken away deliberately. Because they are being replaced by more convenient substitutes that do not produce the same specific meaning. AI-generated art is one instance of this pattern. It is not the only instance.
Which means that anyone concerned about meaning in the current environment has to consider carefully which specific practices to preserve. Making art, in the specific way that requires specific human engagement, is one candidate. Other candidates include specific forms of writing, specific kinds of physical work, specific kinds of interpersonal relationship, specific kinds of care work. All of these can be maintained or automated. All of them will be automated by default unless specific decisions are made to preserve them.
The specific choice to preserve a practice, and to invest in it beyond its economic productivity, is a specific act of maintenance for specific human capacities. It is not sentimentality. It is a specific bet that these capacities are valuable enough to preserve at some cost. And the specific cost is often low, in the long run, compared to what is lost when the capacity atrophies.
Let me name the specific position I have arrived at.
Art in the current moment is a specific form of resistance. Not resistance to AI specifically, which is neither useful nor accurate. Resistance to the specific broader pattern of automation replacing practices that produced human meaning without replacing the meaning itself. Making art directly, in ways that AI could produce more efficiently but that the maker chooses to do anyway, is a specific act of insisting that the practice matters, that the specific human engagement is worth preserving, and that the specific meaning the practice produces is not fully substitutable by outputs alone.
This is not a moral position about AI. AI is a tool, and using it well is worth doing. It is a specific position about what practices to preserve alongside the tools. Making art is one of them. Cooking is another. Building physical things is another. Writing without models is another. Reading long books is another. Each of these is a specific way of preserving a specific human capacity that the broader environment is eroding.
I do all of these things partly as work and partly as maintenance. The specific FALSE GOD project is one of them. Most of what I write is another. The specific way I run my company involves preserving specific human practices alongside the AI tools we use. None of this is done because I think AI is bad. It is done because I think the specific human practices are worth preserving on their own merits, and preservation requires specific ongoing effort.
There is a specific philosophical claim I want to make explicit.
The specific value of art is not fully reducible to the value of the specific artifacts produced. It includes the specific value of the practice for the person doing it, the specific value of the community of practice that develops around specific traditions of making, and the specific value of the historical record that human-made artifacts constitute. Each of these values is real. Each of them depends on specific human practice. None of them transfers to AI-generated outputs.
Which means that anyone claiming AI has replaced or will replace art is making a mistake about what art is. The specific things AI can do involve producing artifacts that look like art. This is not the same thing as doing art. And the confusion of these two things leads to specific bad decisions about which practices to preserve and which to automate.
This is a common mistake in discussions of technology. The specific outputs of a practice get identified with the practice itself, and any technology that can produce those outputs gets taken as a substitute for the practice. This is almost always wrong. The specific practices have specific value beyond their outputs, and the specific value cannot be reproduced by production technology alone.
The specific implication is that the current cultural moment requires specific decisions about which practices to preserve. Making art is one of them. It is worth preserving not because AI cannot produce art-like outputs, but because the specific practice of making art is valuable for reasons that AI-generated outputs cannot replicate. And the specific value can only be preserved by continuing to do the practice, in the specific way that requires specific human engagement.
I want to close with something specific about how this connects to everything else I have been writing.
The pieces I have been publishing about consciousness, about non-human intelligence, about the attention economy, about the trajectories of AI, are all pieces of a specific larger project. The project is to think as clearly as I can about what specifically is happening to human meaning under the specific conditions of the current transitional moment. Some of the writing is about diagnosis. Some of it is about prescription. All of it is about the same specific question, which is how to preserve specific human capacities in an environment that is systematically eroding them.
Art is one of the specific answers to that question. Not the only answer. But a central one. The specific practice of making art, in the specific way that requires specific engagement, preserves specific things that would otherwise be lost. Which is why I do it. Which is why I recommend it, in some form, to anyone who is worried about the specific direction of the current moment.
This does not mean everyone should paint. It means everyone who is worried about the specific pattern I have been describing should find their own specific practice that requires the same specific engagement. Something that resists optimization. Something that produces meaning through the making. Something that develops specific human capacities. Something that leaves a specific human trace in the specific world.
Next month I want to write about the sacred in the digital age, because the specific practices I have been describing point at something that religion used to name and that we have not developed adequate contemporary vocabulary for. The specific relationship between practice, meaning, and the sacred is worth taking seriously as we try to figure out what to preserve. Stay with me.
— Transmission Sent —
Niklas Hanitsch
Reference materials
- David Bayles and Ted Orland — Art & Fear: Observations on the Perils and Rewards of Artmaking (1993)
- Robert Motherwell — The Collected Writings of Robert Motherwell (1992)
- Rick Rubin — The Creative Act: A Way of Being (2023)
- John Berger — Ways of Seeing (1972)
- Boris Groys — In the Flow (2016)
- Byung-Chul Han — The Disappearance of Rituals (2020)
- https://www.newyorker.com/magazine/2022/03/28/why-does-ai-art-look-like-that
- https://www.moma.org/collection/works
Continue reading
- Dark aesthetics as tech critique: The visual language of digital dread
- Where does the sacred live in the digital age
- Ritual is not superstition. It is technology for the soul
- The Purpose Behind FALSE GOD
- Alone with My Own Hunger: My First Solo Exhibition
Frequently asked questions
Can AI make real art? AI can produce artifacts that look like art and that can be aesthetically compelling. Whether these count as art depends on how you define art. If art is defined by the specific practice of making meaning through engaged human production, AI outputs are not art in that sense. If art is defined by the aesthetic qualities of the resulting artifact, AI outputs can be art. The definition matters.
Why does the practice of making matter if the output looks the same? Because art is not only the output. The practice of making produces specific psychological transformation, develops specific capacities, and creates specific artifacts that carry information about the maker's engagement. None of these transfer to AI outputs, even when the outputs are aesthetically equivalent to human-made artifacts.
Should artists use AI tools? That depends on the artist. AI tools can be useful for specific purposes and are worth using well. What matters is not whether AI is used but whether the specific practice of engaged making is preserved alongside AI use. Artists who use AI as one tool among many can maintain the practice. Artists who use AI to replace the practice lose what art was doing for them.
Will AI replace human artists? For some markets, partly. AI will replace human artists for certain kinds of production where the outputs are the only thing that matters and where volume and speed are more valuable than specific human engagement. It will not replace human artists for contexts where the specific practice of making matters. Both markets will continue to exist.
Why is art particularly important in the age of AI? Because art is one of the specific practices that requires the specific human capacities the broader environment is eroding. Preserving the practice of making art is a specific way of preserving those capacities, and by extension of preserving the specific kinds of meaning that depend on them. This makes art unusually important in the current moment, not less important.