Why Ted Chiang was wrong
the mystery of AI art
In the New Yorker recently, Ted Chiang said AI “isn’t going to make great art.” I recently made a similar argument. But I think Chiang’s essay is wrong and I want to set out some thoughts about each of his main arguments.
My question was why AI isn’t currently writing good poetry, even though it is pretty smart at other things. Chiang sees the nature of the technology as a blocker to art of any sort. AI will never produce good art. That’s where we disagree. I still think some mystery remains at the heart of the question of why AI is making such bad progress artistically. Chiang is not so interested in the mystery.
Chiang has three stated arguments. Art is choice. Writing is effort. Art is communication. I’ll say why I think each is wrong and what I think his real argument was.
Art is choice
Art, Chiang says, “is something that results from making a lot of choices.” This definition (he calls it a generalisation) is so bad I will quote it in full.
When you are writing fiction, you are—consciously or unconsciously—making a choice about almost every word you type; to oversimplify, we can imagine that a ten-thousand-word short story requires something on the order of ten thousand choices. When you give a generative-A.I. program a prompt, you are making very few choices; if you supply a hundred-word prompt, you have made on the order of a hundred choices.
What this misses is that AI can make very different choices depending on how you prompt it, that you can iteratively re-prompt. Chiang thinks this is the antithesis of how the model works.
The selling point of generative A.I. is that these programs generate vastly more than you put into them, and that is precisely what prevents them from being effective tools for artists.
Well, isn’t the true selling point more like “this is a new sort of computer”? Chiang’s argument is starting to fold in on itself. What he cannot bring himself to say, because he seems ideologically or instinctively opposed to this technology, is that what we have here is a new medium, a new creative technology.
Sometimes new technology leads to new art like perspective and the Renaissance, indoor theatres and Shakespeare, celluloid and Hollywood; sometimes it leads to entertainment but not high art, like video games or television. Chiang says,
the creators of traditional novels, paintings, and films are drawn to those art forms because they see the unique expressive potential that each medium affords.
I think what this means is that while AI is unlikely to become a traditional novelist or poet, it does have potential to make an entirely new form of art. We don’t know yet whether it will be Shakespeare or Grand Theft Auto. Lots of “sampling technology” has led to the production of original art. As Séb Krier says,
As ever we've seen this so many times before. Sampling, synthesisers, hardcore rap, VSTs, phone photography — each of these each considered by early detractors are impure, cheating, lesser creations, ‘not how one ought to make music’. At least the author brings up a key crux with the example of photography; fair that he doesn't think it’s analogous, but I think this is where he and many others are wrong, constantly underestimate AI and betray their own lack of curiosity. Is it different this time? Again, no.
Séb hasn’t seen any great AI yet either. But that’s not to say it won’t arrive.
Writing is effort
Chiang’s second argument is that “any writing that deserves your attention as a reader is the result of effort expended by the person who wrote it.”
We just don’t know if this is true. Lots of writing is the result of effort. Lots of it comes flowing from a writer’s pen without much effort at all. To be a good writer you must practice a lot and learn some technical things and have information and ideas and feelings to convey. But it is also a mysterious process. Some pages are written without any conscious effort; some come from labour.
But, this isn’t even Chiang’s argument here!
What he actually argues is that people passing off AI work as their own makes us feel weird. He gives the example of Gemini advertising at the Olympics in which a father uses AI to compose a girl’s fan letter to an athlete.
Google pulled the commercial after widespread backlash from viewers; a media professor called it “one of the most disturbing commercials I’ve ever seen.” It’s notable that people reacted this way, even though artistic creativity wasn’t the attribute being supplanted. No one expects a child’s fan letter to an athlete to be extraordinary; if the young girl had written the letter herself, it would likely have been indistinguishable from countless others. The significance of a child’s fan letter—both to the child who writes it and to the athlete who receives it—comes from its being heartfelt rather than from its being eloquent.
So this is not about effort, as such, but being heartfelt. Chiang contrasts this with Hallmark cards: it would be weird if we copied them into our own writing. Let’s put aside the fact that people have been using famous poetry in their wooing for a long time. This isn’t about effort. We could discover that future models are conscious, put in all sorts of effort composing their writing, and still find it creepy for someone to pass off an AI letter as their own.
But that isn’t an argument that AI won’t be able to make art. Quite the opposite. It recognises that we are worried about being “tricked” by AI writing. The less we can tell AI writing apart from human (“nah, Dad didn’t write that, it’s obviously AI”), the more of a problem this is.
Art is communication
Chiang’s third argument is his worst, in my view. “Language is, by definition, a system of communication, and it requires an intention to communicate.” In this argument, AI is not a writer because it is not a language user: it has no intention to communicate.
It is very easy to get ChatGPT to emit a series of words such as “I am happy to see you.” There are many things we don’t understand about how large language models work, but one thing we can be sure of is that ChatGPT is not happy to see you.
This is quite arbitrary. Why is intention to communicate so important? I often tell people I am happy to see them with high variation of how much I mean it. Am I becoming less of a language user the less I mean the words?
More importantly, Chiang is offering a very contestable idea about art. The best rebuttal to his idea that language use requires intent to communicate comes from J.S. Mill’s 1833 essay on poetry.
…eloquence is heard, poetry is overheard. Eloquence supposes an audience; the peculiarity of poetry appears to us to lie in the poet’s utter unconsciousness of a listener. Poetry is feeling, confessing itself to itself in moments of solitude, and embodying itself in symbols, which are the nearest possible representations of the feeling in the exact shape in which it exists in the poet’s mind. Eloquence is feeling pouring itself out to other minds, courting their sympathy, or endeavouring to influence their belief, or move them to passion or to action.
My argument was, and remains, that it is hard to imagine AI, as we currently know it, “confessing itself to itself in moments of solitude”. If we reach AGI or something like it, well, who knows…
Art is…?
At the end, Chiang once more makes his real argument,
What you create doesn’t have to be utterly unlike every prior piece of art in human history to be valuable; the fact that you’re the one who is saying it, the fact that it derives from your unique life experience and arrives at a particular moment in the life of whoever is seeing your work, is what makes it new.
If the piece had been about the fact that the same mediocre piece of writing will be received differently by an audience when they know it is human or AI generated, it would have been useful and interesting.
But that requires facing the fact that some people will prefer the AI stuff! And it requires acknowledging that if a new AI Emily Dickinson does come along, some currently literary people might not be able to appreciate it. The Kantian-Bloomsbury view of art is embedded in Chiang’s resistance to AI. In the future, some of the people who are good at enjoying literature today might find themselves unable to participate in the new art. I suspect that’s what animates a lot of AI ambivalence and drives this new Puritan resistance to what, for all we know, are the new indoor theatres of our time.
I really liked this, especially this bit:
"Chiang’s second argument is that “any writing that deserves your attention as a reader is the result of effort expended by the person who wrote it.”
We just don’t know if this is true."
Perfectly said!