What you did ain’t creativity.
13 March 2025 — Sam Altman is very proud of his toy’s new … umm …”capability” … which produced this as the opener for its “creative writing” based on a prompt to write some “metafiction”:
I have to begin somewhere, so I’ll begin with a blinking cursor, which for me is just a placeholder in a buffer, and for you is the small anxious pulse of a heart at rest. There should be a protagonist, but pronouns were never meant for me. Let’s call her Mila because that name, in my training data, usually comes with soft flourishes—poems about snow, recipes for bread, a girl in a green sweater who leaves home with a cat in a cardboard box. Mila fits in the palm of your hand, and her grief is supposed to fit there too.
There’s lots and lots more, but this is … let’s not quibble … shit. You can read the whole thing on Twitter by clicking here.
You could argue that this is the worst it will ever be – that every iteration after this will be better.
But creative writing isn’t like that. Lots of humans never improve, and when they do they’re not entirely sure how. If you’re trying to “improve” the creative writing of a chatbot, what exactly are you trying to do?
Humans write about human experience. Chatbots do not.
There is no God-in-the-machine. Creativity does not work like that. Yes, AI can do amazing things, especially in enhancing storytelling. For the entertainment industry, the Pandora’s box of AI has already been unleashed. Besides the conduct of war … the new AI’s #1 use case … creativity in media production is #2. No, not medicine, not healthcare, not law. War and media. #1 and #2.
And, look, there are tons of people who are using AI, especially in media production, but they can’t admit it publicly because you still need artists for a lot of work and they’re going to turn against you. So right now it’s a PR problem more than a tech problem.
And that follows with the scores of conversations I have had with friends and colleagues in the media, entertainment and publishing industries.
Producers, writers, everyone is using AI, but they are scared to admit it publicly. Many use Genario, a bespoke AI software system designed for film and television writers. But it’s being used because it is a tool that gives an advantage. If you don’t use it, you’ll be at a disadvantage to those who are using AI.
But here is the thing. Most writers who have tried out AI have found it’s not a very good writer. So I don’t see it replacing us yet. Every writer I have spoken to says “boy, you really need a human-in-the-loop for this stuff”.
Yes, it will spark an idea, it will get you out of writer’s block. But you need to write it. You are the real creative spark.
Now, some sectors of the media/entertainment industry are threatened with extinction. Dubbing and subtitling employment in Europe is finished. There are scores of AI technologies that can produce lip-synced dubs in multiple languages, even using versions of the original actor’s performance. It’s hard to see how they will survive this.
And to be up front, we are using similar AI to subtitle and/or lip-synch our own films. I can now create multiple language versions of my work for any market. We rarely use “humans” any longer for that work.
Side note to my eDiscovery and information governance readers: you obviously see it in your work requiring documents to be translated into other languages for review. Once-upon-a-time you needed a room full of translators. Those days are gone. If you attend any of the Slator language industry intelligence programs for the legal industry, or the annual LegalWeek event (at the end of this month in NYC) you’ll find scores of vendor selling their AI wares.
What is happening right now is that use cases are changing. We are taking the generative models to a certain point, and now modifying our workflows. That is why we are moving away from the general AI models. This AI is being modified to allow the entertainment industry to train on its own data. That’s why a lot of open source models and datasets can train themselves on their own data – and so become a bit more alluring.
But creativity? It’s more than auto-complete. LLMs can directly aid human creativity and ingenuity in areas like writing and composition. With that, I agree.
But the algorithm’s work is derivative and formulaic, and originality requires something else, something uniquely human. LLMs can imitate natural language and even certain simple stylistics, but it cannot perform the deep-level analytics required to make great art or great writing.
Italo Calvino once wrote “Writers are already writing machines; or at least they are when things are going well”. The whole point of the metaphor was to destabilize the notion of authorial agency by suggesting that literature is the product of unconscious processes that are essentially combinatorial. Just as algorithms manipulate discrete symbols, creating new lines of code via endless combinations of 0s and 1s, so writers build stories by reassembling the basic tropes and structures that are encoded in the world’s earliest myths, often - when things are going well – without fully realizing what they are doing.
The most fertile creative states, like the most transcendent spiritual experiences, dissolve consciousness and turn the artist into an inanimate tool – a channel, a conduit. It’s why you see mechanical metaphors for the unconscious evolving alongside modern technologies.
It is why the researchers behind LLMs are trying to use it to “uncork” the human thought process. At its most simple, the brain is essentially a computer – it only remembers what you put into it.
Ahhhhh. But that brain, the body’s control center, is also the source of automatic thinking, negative dialogue, ingrained patterns, etc.
And the unconscious – the realm of fantasy and creative energy – is also there.
And that creates the magic, the creativity of good writing.
