Right now, a lot of artists and creatives are angry about AI.
Alarm bells are sounding in the arts and creative industries, with fears of human creators being replaced by AI, resulting in job losses, companies failing, and even whole industry sectors finding themselves redundant.
To take a few recent headlines:
Graphic design among most at-risk jobs from AI – report
Design WeekHuman artists could disappear if copyright not protected from AI, MPs told
The IndependentWPP’s Mark Read Says AI Will Upend the Advertising Workforce
AdweekAI plundering scripts poses ‘direct threat’ to UK screen sector, says BFI
The Guardian
All of this is generating not only anxiety among creatives, but also anger, given the allegations of AI models having been trained on creators’ work without their permission.
And yet… a lot of the enthusiasm about AI creativity is coming from technologists who, with the greatest respect, may be very good at what we could call ‘technical creativity’, but don’t really understand artistic creativity.
And the alarm bells are often triggered by journalists and writers (or their editors) using sensational headlines that obscure the real story.
Just to be clear: I’m aware that some artists are making brilliant use of AI in their creative process. But given all the hype about AI threatening to ‘replace’ human creators, I’m focusing on a narrower questions: can autonomous AI, working without human creative intent, produce art?
Apparently, AI is on the verge of mastering poetry
I’ll take my own art, poetry, as an example. You can imagine my scepticism on seeing the following headline:
This AI Poet Mastered Rhythm, Rhyme, and Natural Language to Write Like Shakespeare
The article is a lot more thoughtful and nuanced than the headline, and makes it clear that this AI poet did not master rhythm, rhyme and natural language to write like Shakespeare.
Here’s another example:
AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably
This is the title of a fascinating research paper by Brian Porter and Edouard Machery of the University of Pittsburgh, published on Nature.com.
The paper has been widely reported, with attention-grabbing headlines such as:
ChatGPT is a poet. A new study shows people prefer its verses
The Washington PostAI poetry rated better than poems written by humans, study shows
The GuardianArtificial intelligence is better at writing poems than William Shakespeare, a bizarre new study has found
The Scottish Sun
It has also been widely discussed on forum threads such as ‘Has AI Killed Poetry?’ (Reddit).
And its findings make sobering reading for those of us who cherish the idea that poetry is a unique and special form of human artistic expression.
In the researchers’ first experiment, they gave 1,634 non-expert readers 10 poems – 5 written by poets such as Chaucer, Shakespeare, T. S. Eliot, Sylvia Plath and Dorothea Lasky, and 5 AI imitations of the same poets. The readers were asked to decide whether they thought each poem was written by an AI or a human.
Overall, readers rated the AI-generated poems as more likely to be human-authored than the actual human-authored poems.
In a second experiment, they asked participants to rate the quality of the poems. And the readers thought the AI poems were better than the works of distinguished human poets.
Better in what way? Specifically, they thought the AI poems were more moving, profound, lyrical, inspiring, beautiful and meaningful than the human poems. When it came to originality, the AI poets were judged to be equally as original as the human (ahem) originals.
I must admit my heart sank when I read this. ‘Maybe the robots really are coming for poets,’ I thought.
One thing that is missing from the research report is examples of the AI poems that were judged to be superior to the human efforts. Fortunately, ChatGPT came to my rescue, and found me some examples in an article about the research on the Smithsonian website.
So here for your consideration is an AI attempt at emulating a poem by Sylvia Plath:
The air is thick with tension,
My mind a tangled mess.
The weight of my emotions
Is heavy on my chest.The darkness creeps upon me,
A suffocating cloak.
The world outside is cruel and cold,
And I’m a fragile, broken yolk.My thoughts are spinning wildly,
A cyclone in my brain.
I try to grasp at something solid,
But all is lost in vain.The voices in my head,
They never cease to scream.
And though I try to shut them out,
They haunt me like a dream.So here I am, alone and lost,
A ship without a sail.
In this world of pain and sorrow,
I am but a mere wail.
What do you think? Is this is more moving, profound, lyrical, inspiring, beautiful and meaningful than the actual poems written by Sylvia Plath?
(If you’re not familiar with her work, here’s a selection of poems by Plath on the US Poetry Foundation website.)
If you want to know my opinion…
The AI Plath is truly terrible – clunky, clichéd and tin-eared. It’s not just bad, it’s laughably bad. It is also nothing remotely like Plath’s poetry.
And given Plath’s well-documented struggles with mental health, lines such as ‘The voices in my head, / They never cease to scream.’ are crass and insensitive. A human poet would be much less likely to make this mistake, but of course, the AI doesn’t know what it’s writing.
If this really were the competition, poets would have absolutely nothing to fear from AI.
Yes, it’s still depressing that so many readers thought this was better than the real Sylvia Plath. But I humbly beg to differ. And Plath has so many admirers that I don’t think we need to worry about her artistic legacy.
There is of course better AI poetry out there than this. And Porter and Machery have conducted a thoughtful and thought-provoking study. But this example shows we should look beyond the headlines when it comes to assessing bold claims about AI creativity.
I’m not just talking about technique or originality
My reaction to AI poetry is similar to that of a lot of creatives when they see AI versions of their discipline: the AI products look crude, artificial and amateurish. To us, they are interesting novelties, remarkable in their own right, but the shortcomings are obvious, even if others are wowed.
But I don’t think we should rest on our laurels, at least from a technical viewpoint. AI expert Ethan Mollick advises us to assume that today’s AI is ‘the worst you will ever use’. The technology will only get better. So we can expect its creative technique to keep improving.
One of Mollick’s best-known articles is The recent history of AI in 32 otters, where he fed the same prompt (‘otter on a plane using wifi’) into Midjourney between 2022 and 2025, and published the images it produced. Looking at the succession of images, it’s clear that Midjourney is getting better and better at creating images. So it seems conceivable to me that AI could end up creating work that was technically indistinguishable from the work of the best human illustrators.
Maybe AI is better at illustration than poetry. But I still chose an image by a human illustrator for this article. Maybe one day, AIs poetic technique could improve enough to imitate really good poetry, to a standard that would fool even expert readers.
But it still wouldn’t be poetry. Just as even a flawlessly-executed AI image or sculpture or film or piece of music wouldn’t be art.
Another common criticism of AI creativity is that its productions are essentially aggregates of the data from its training, so that it tends towards bland imitation rather than striking originality.
But once again, I’m going to grant the technologists the benefit of the doubt. I’m going to accept that AI is an evolving field, and that future AIs will be able to think more creatively.
In this imagined future, AI will be able to create artefacts – text, images, video, sound, etc. – that are as original as any human work, and indistinguishable in terms of their technical execution. Even experts are fooled by its creations.
That would be amazing, but it still wouldn’t be art.
Art is a meeting of minds (and AI is mindless)
The reason these strikingly-original and perfectly-executed artefacts would not be art is that they would be missing one crucial thing that is essential to real art – the sense of having been created by a living, breathing, feeling, suffering human being.
When we read a poem by Sylvia Plath, part of what makes it meaningful to us is the knowledge that it was written out of a fiercely intense lived experience.
The same is true of any great poem. It matters that it were written by someone who has lived and loved and experienced the kind of thoughts and feelings it expresses – even if the poetry isn’t straightforwardly autobiographical or even realistic.
Poetry, like any art, is not just a technical performance. It is a meeting of hearts and minds.
We value art for its ability to communicate more than information, to convey something essential about the human experience.
Most of the time, we take this relationship between artist and audience for granted. It’s so obvious that it doesn’t require comment. But it becomes glaringly obvious when the unspoken contract is broken.
In his book The Creativity Code, Marcus du Sautoy recounts a series of daring experiments conducted by David Cope and Douglas Hofstadter, where they staged concerts featuring unfamiliar pieces by Bach and Chopin, as well as imitations of these composers by Cope’s EMI (Experiments in Musical Intelligence) software. The audiences were made up of classical music lovers, including composers and music theorists.
To everyone’s shock, a piece by the real Bach was judged ‘a poor imitation’ of EMI’s imitation. The same thing happened with Chopin.
One audience member described the audience’s reaction on learning they had voted the wrong way:
There was a collective gasp and… what I can only describe as delighted horror.
Some listeners were so incensed they vented their fury on Hofstadter, to the point where he felt physically threatened after one concert.
Audiences hate being fooled like this, because it suggests they are not as discerning as they thought they were. More than that, they hate the idea that the profoundly moving experience they just had was based on a fake. It feels like deception, even betrayal.
Does this example show how easily artistic types can be tricked? That we are not as sophisticated as we like to think? Maybe there’s a bit of that. But it’s not the whole story, because art has never been a one-way street – in spite of the popular image of the creative genius expressing their vision for the rest of us to marvel at.
Without a thoughtful and engaged viewer, a Picasso is just an old canvas covered in paint. Without a sensitive reader, a novel by Dickens is just a succession of marks on a page.
Art requires at least two people to engage in a relationship, via the medium of the artefact. And like any relationship, it breaks down if one person turns out to be a fake.
Imagine chatting online to someone who you thought was a human being, a potential friend or romantic partner, only to discover it was a bot. Most people would lose interest at this point. Just as the AI Bach lost its allure the moment the audience discovered it was generated by a software program.
To return to Porter and Machary’s research paper, they cite an impressive range of studies showing AI creative work becoming ‘indistinguishable’ from the human versions – but there is an important caveat (emphasis mine):
AI-generated images have become indistinguishable from reality. AI-generated paintings are judged to be human-created artworks at higher rates than actual human-created paintings; AI-generated faces are judged to be real human faces at higher rate than actual photos of human faces, and AI-generated humor is just as funny as human-generated jokes. Despite this, studies have consistently found a bias against AI-generated artwork; when told that an artwork is AI-generated, participants rate the work as lower quality.
Even when AI creative work is rated better in a blind test, as soon as the true origin is revealed, audiences recoil from it. Because we don’t just want polished perfection in art: we want an authentic human connection.
And AI cannot provide, this, because however good it gets at simulating the patterns and passions of human art, it can never feel those passions itself.
It’s good at what Alan Turing called ‘the imitation game’, but not the real thing.
Because it doesn’t know what it’s like to live and move and breathe, or to feel emotions at the visceral level.
It has never been head-over-heels in love, or sick to the pit of its stomach.
It has never savoured a delicious meal in entrancing company. It has never had mind-blowing sex.
It has never woken from a profound sleep to marvel afresh at the miracle of life, listening to the dawn chorus greet the new day.
It has never woken from troubled dreams to suddenly remember the loss of a loved one, and felt a fresh wave of grief.
And it does not know that it will die.
And this is something we value in poetry and art in general: it does come from a living breathing, loving, suffering, mortal human being.
So however good it gets at technique, however original its ideas may be, however well it simulates emotion and experience, AI cannot make real art. And it probably never will.
Art is a game of meaning (and AI is meaningless)
To be fair to AI experts, they are very well aware of this issue. It just seems to have been overlooked in the popular hype around ‘AI creativity’.
In 1980 the philosopher John Searle kicked off a long-running debate in his paper ‘Minds, Brains, and Programs’, where he advanced what is known as the Chinese room argument.
He imagined a person with no knowledge of the Chinese language sitting isolated in a room. The room contains a book of instructions for manipulating Chinese characters. Every so often a Chinese text is delivered to the room, and the person follows the instructions in the manual to produce texts that read to Chinese speakers as meaningful responses. The person in the room, however, is just following a set of rules, and has no idea what any of the Chinese texts mean.
Searle’s contention is that computers executing programs are like the person in the Chinese room: they are just processors, with no understanding of the meaning of what they are processing, and no mind or consciousness comparable to that of humans.
It’s an ongoing debate among philosophers, technologists and other specialists. But for my purposes here we don’t need to resolve the philosophical question. We just need to consider the fact that, throughout history, humans have taken it for granted that art is produced by other humans, with a mind and consciousness like their own.
This is why many writers on creativity and cultural theory, such as Mihaly Csikszentmihalyi, Howard Becker and Pierre Bourdieu, have described creativity not as the work of a lone genius but as a social process – an ongoing conversation between creators, gatekeepers and audiences.
A canvas and paint have no intrinsic meaning, only the meanings we ascribe to them. The same artwork can have different meanings for different people or groups, in different contexts. So creativity becomes a game of meaning, with many players.
Following Searle’s logic, if AI only shuffles symbols without truly understanding them, then an AI artist does not have a mind like humans. So it cannot participate in these networks and relationships the way we do. It cannot play the game of meaning.
When we encounter a new artwork, questions arise in our mind:
What does it mean?
Did the artist really mean that?
Am I missing something here?
So the game of meaning involves ambiguity, subjectivity and interpretation. It results in a process of negotiation between artists, critics and the public, about what constitutes art.
And it turns out that this aspect of art doesn’t require any executional skill at all.
When Duchamp put a urinal in the 1917 Society of Independent Artists exhibition in New York, and labelled it ‘Fountain’, he made a bold move in the game of meaning. He didn’t create the artefact, all he did was reframe it as art, by giving it a title and a place in the exhibition.
Tracy Emin made a similar move when she exhibited her unmade bed in the Tate Gallery in 1999 and entered it in the Turner Prize.
Both works have come to be accepted as genuine art, with ‘Fountain’ being voted the most influential artwork of the 20th century by 500 art professionals, and ‘My Bed’ being sold at Christie’s for £2,546,500.
Yet initially, both works provoked squawks of outrage. And there have been plenty of people who tried to persuade the rest of us that their work was art, and the rest of us remained unpersuaded.
‘Call that art?’ some people ask, when confronted with something new.
‘Yes,’ says the artist. ‘I do call it art.’
And we argue and decide whether or not this ‘counts’, whether it’s a valid move in the great game of art that artists and audiences have been playing for thousands of years.
But because AI lacks consciousness, because meaning has no meaning for AI, it is not currently a player in this game.
About that ‘probably’…
In the title of this article I say that AI will ‘probably’ never create art. But there are no guarantees. It is conceivable that our concept of art, and/or our attitudes to AI, will shift to the point where AI art is considered as valid and meaningful as human art.
There are two obvious scenarios in which this could happen.
Scenario 1. We cease to see art as a meeting of minds
In this scenario, our definition of art, and the kind of experience we want from it, evolves with our technology.
Instead of seeing art as a meeting of minds, we value it as a mirror of our own consciousness. We accept that there is no ‘other’ behind AI art, no second player in the game of meaning.
Instead, all the meaning we find in the art is generated by the viewer. When we contemplate such art, we are using it to reflect on our own thoughts, feelings, sensations and experiences.
So this art of the future, is more of a single player game, rather than the multi-player game we join when we gaze at a Rubens or watch a Tennessee Williams play or listen to Mozart.
Personally, this would stretch my own definition of art to breaking point. For me, it would no longer be the real thing. But I’m on the far side of fifty, and my vote won’t count in the long run. Maybe future generations will think differently about all this. It will be for them to decide.
Scenario 2. AI becomes (or turns out to be) conscious
This second scenario is more startling and more revolutionary. In this future, we come to accept that AI has become conscious. Maybe we even decide that the AIs of the 2020s were already conscious, if only in a limited way.
In this future, AI art becomes a meeting of very different minds. A bridge between AI consciousness and ours.
In 2018 I interviewed the artist Tyler Hobbs for my podcast, and asked him whether he thought AI could ever creat art that we find as moving as human art. Here’s his reply:
Let’s imagine a future where an AI lives in a robot and it can go around and collect its own physical images and sensory input and it’s working based on that. I think it’s going to be really tough for an AI to produce artwork if we define artwork as being… about the human experience in some way, then just from that definition, I don’t think an AI is going to be able to create truly new artwork, just because it doesn’t experience and won’t experience the world and life in the same way that a human does.
But maybe the AI can very successfully create artwork that is meaningful to itself or to other similar AI. And in that case, it might be a very successful artist. But yeah, it’s going to be really interesting if this stuff develops to see what an AI’s idea of artwork is once they can move beyond mimicry as the main operation. I have no idea what that’s going to look like or how successful it might be. But I think it’ll give us some interesting insight into ourselves and how AI might fundamentally be different from humans.
This would arguably be real art and a whole new ballgame. It would open up hitherto unimaginable realms of art.
If we ever find ourselves in a world of AIs that we consider conscious, then maybe art could be a tiny thread that connects us with these superintelligent beings, and permits some kind of meaningful communication with them.
This art could perform a similar function to the logographic language in Ted Chiang’s short story ‘Story of Your Life’ (adapted into the movie Arrival) that enables humans to communicate with heptapod alien visitors. Or the musical notes that form a bridge between the humans and aliens in Spielberg’s Close Encounters of the Third Kind.
It feels odd to be even writing about this kind of scenario. It doesn’t feel likely. But then it wouldn’t, given that it involves trying to imagine what happens beyond our current conceptual horizon.
Whatever you make, make it art
To restate my basic point: as long as we see art as a meeting of minds, and as long as AI is not conscious, then AI cannot make art.
Yet not all creative work is art. A lot of it is pretty average or disposable, or merely functional. And the more mediocre or functional it is, the easier it can be automated.
AI is already doing some types of advertising, design and decorative work. No one really cares if a pay-per-click ad was written by a robot.
Maybe AI will produce dazzling entertainment, in the form of TV, movies, and serial fiction, and some audiences will be happy to consume it.
So the creative sector is clearly in for major disruption. None of us can be complacent. But the closer your work approaches the condition of art – where human experience and human connection matter – the harder it will be for an AI to compete with you, let alone replace you.
So whatever you create from this point, remember that AI is gobbling up all the low-handing fruit. There’s no safety in doing something mediocre or even pretty good.
Whether or not you have previously considered yourself an artist, whether you work in advertising, TV, gaming, design, or another commercial creative field, ask yourself:
How can I take my work to the level of art?
How can I channel my unique human experience into my work?
What kind of meaningful connection do I want to make with others?
