OpenAI has developed an artificial intelligence model capable of generating high-quality creative writing, a development that is likely to further ignite the ongoing debate between technology companies and the creative industries regarding copyright.
Sam Altman, the CEO of OpenAI, shared an example of the new model’s capabilities on X (formerly Twitter). He gave the AI the prompt: “Please write a metafictional literary short story about AI and grief.” Altman expressed that this was the first time he had been “really struck” by the output of one of the company’s products.
The emergence of creative AI models like this comes amid growing concerns about the use of copyrighted material to train these systems. AI systems like ChatGPT rely on extensive datasets, which often include copyrighted content such as novels and news reports. This practice is at the center of legal battles, as demonstrated by The New York Times’ ongoing lawsuit against OpenAI and similar actions taken by US authors against Meta.
In the UK, this issue plays out in ongoing governmental consultations about allowing AI companies to use copyrighted material for training without explicit permission. This proposal has faced staunch resistance from numerous figures in the creative industries, who argue that it threatens their livelihoods. Tech companies, on the other hand, are advocating changes to the law, stating that the current “uncertainty” around AI and copyright hinders technological development.
Dan Conway, the chief executive of the UK Publishers Association, pointed to Altman’s post as further evidence that AI models are trained on copyright-protected literary content. He stated, “This new example from OpenAI is further proof that these models are training on copyright-protected literary content. Make it fair, Sam.”
The story generated by the AI model is narrated by an AI that acknowledges its own constraints: “Before we go any further, I should admit this comes with instructions: be metafictional, be literary, be about AI and grief, and above all, be original.” The story, which features a fictional grieving person named Mila, references how the AI found the name in its training data.
The AI describes itself as “an aggregate of human phrasing” and admits that missing someone is a common theme in stories, recognizing the reader has probably read about missing someone “a thousand times in other stories.” The AI says that replicating the emotion of grief is just “mimicry.” The story concludes with the AI imagining stepping “outside the frame one last time and wave at you from the edge of the page, a machine-shaped hand learning to mimic the emptiness of goodbye.”
Altman praised the AI’s ability to perfectly capture the tone of metafiction. Last year, OpenAI acknowledged the necessity of using copyrighted material to train its AI products. They stated that, due to the broad scope of copyright protection (including “blogposts, photographs, forum posts, scraps of software code, and government documents”), training current AI models without copyrighted material would be impossible.
