Generative Art Before AI: What Tate Says About Machine Creativity

Whenever a new AI image tool goes viral, the same argument returns: is this really art, and who made it? It feels brand new. But the idea of handing part of the creative process to a system is much older than today's apps. A good place to start is a short, clear definition from the Tate museum, and it changes how I read the whole debate.

What Tate means by generative art

In its glossary, Tate defines generative art as art made using a predetermined system that often includes an element of chance. It adds that the term is usually applied to computer-based art. That is a modest definition, and it is useful for that reason.

Notice what it does not say. It does not say the art must come from a neural network, or that a human must be absent. The key ingredients are a system set up in advance and, often, a dose of randomness. The artist designs the rules, and the result is partly unpredictable.

This definition is a handy yardstick. Many of today's AI image generators fit the broad pattern: a system built beforehand, a person giving it an input, and an output that is never exactly the same twice. I should be careful, though. Tate's page does not discuss modern AI image models, so what follows about them is my own reading, not a claim from the museum.

A longer history than the headlines suggest

Tate traces the practice back to dada, the art movement that, in Tate's own glossary, formed during the First World War in Zurich. Dada is closely tied to the embrace of chance and the rejection of tidy tradition. Seeing generative art linked to it is a reminder that artists have long been curious about letting something other than their own hand steer the outcome.

Tate then names a pioneer: Harold Cohen, considered one of the first practitioners of generative art. According to Tate, he used computer-controlled robots to generate paintings in the late 1960s. Tate's page shows one of his works, an untitled computer drawing from 1982. So a person working with a machine to produce images is not a 2020s invention. It is more than half a century old.

Tate also mentions the Turner Prize winner Keith Tyson, who built what it calls an ArtMachine, a complex recursive system that generated detailed propositions for artworks, which Tyson then made. This example is the one I find most interesting. The machine produced instructions, and the human carried them out. The creative chain ran in both directions.

Who is the author when a system is involved?

Here is where the history helps with today's arguments. In Cohen's case and in Tyson's, nobody claimed the machine was the artist alone, and nobody claimed the human did everything. The artist designed or chose the system, and the system surprised the artist. Authorship was shared in a way that was awkward but workable.

With modern AI tools, the question gets harder because the systems are far more general. A person typing a sentence into an image generator contributes much less specific design than someone who writes their own program. And the system itself may have been built from the work of countless other people. I am not going to make claims about specific lawsuits or rulings here, because I have not verified current details, and that area changes fast. But the underlying question is clear: where does the human contribution begin and end?

My own view is that the Tate definition points to a useful test. Ask who designed the system, who chose the inputs, who selected the result, and who took responsibility for showing it to the world. Each of those is a creative act, and the answers differ from piece to piece. The label "AI art" hides that variety.

The role of chance

The phrase "an element of chance" deserves attention. Chance can be liberating: it lets an artist see options they would never have chosen deliberately. But chance alone is not art. It becomes meaningful when someone decides which of the random outcomes to keep and why.

This is why I am skeptical of both extremes in the conversation. One extreme says that AI output is just noise with a glossy finish, and therefore worthless. The other says that anyone who types a prompt is now an artist in the full sense. Tate's definition quietly suggests a middle path. The system and the chance are ingredients. The judgment is what turns them into a work.

Tyson's case, as Tate describes it, shows this nicely. The system generated propositions, but he was the one who made the artworks. The selection and the making stayed with the person.

What this means for critics and audiences

If art made by systems has a history, then criticism of it does too. Critics have long had to ask whether a piece is interesting because of the idea behind it, the system that produced it, or the final image. That question is now being asked by a much wider public, because anyone can try these tools in minutes.

I think this is healthy. It pushes people to look more closely at how images are made, and to be a little more suspicious of anything that looks polished. It also gives critics a new job: not only judging the result, but explaining the process, the choices and the context, so audiences can tell a thoughtful use of a system from a casual one.

If you want to explore more of this on our site, we have looked at the topic before in our piece on AI, art creation and critique and in our article on how AI is reinventing creativity. Those are general overviews, while this one is about a single idea: that the machine as co-creator is an old story.

Related creative fields

The same logic shows up well beyond galleries. Designers use automated tools to generate options and then choose among them, which we covered in our look at AI in graphic design. Game developers use systems to populate worlds, a theme we explored in our article on AI and video game design. In each field the pattern is similar: a system proposes, a person decides.

That pattern is reassuring and unsettling at once. Reassuring, because it suggests human judgment still matters. Unsettling, because the more capable the system becomes, the more the human role risks shrinking to a single click of approval. Whether that happens depends less on the technology than on the habits of the people using it.

Keeping the human in the loop

So what should we take from a short museum glossary entry? First, that generative art is not synonymous with the latest AI craze. It has roots in dada, a pioneer in Harold Cohen, and later experiments such as Keith Tyson's ArtMachine, all according to Tate. Second, that the core of the idea is a system plus chance, with a human who designs, selects or interprets.

Third, and this is my opinion, the interesting question is not whether machines can make art. It is what we want people to keep doing. Choosing, refusing, editing and taking responsibility are creative acts. If we protect them, tools can widen what artists do. If we forget them, we may end up with an endless supply of images and very little meaning.

Next time you see an AI-made image, try asking the questions the history suggests. Who set up the system? What was left to chance? Who decided this was the one to show? And in your own work or hobbies, where would you be happy to let a system surprise you, and where would you insist on keeping the final say?