OpenAI Working on a New Model “Good” at Creative Writing

OpenAI has developed a new AI model that excels at creative writing, according to CEO Sam Altman.

“We trained a new model that is good at creative writing (not sure yet how/when it will get released),” Altman wrote in a post on X.

To showcase its capabilities, he shared a sample of its work – a 1,172-word short story generated in response to the prompt: “Please write a metafictional literary short story about AI and grief.”

The story features a protagonist named Mila, who engages in an extended dialogue with the AI chatbot following the loss of her partner, Kai. Through this narrative, the model showcases its ability to delve into complex emotions and relationships, producing creative and emotionally resonant content.

“This is the first time I have been really struck by something written by AI; it got the vibe of metafiction so right,” Altman enthused.

The announcement sparked mixed reactions. Some critics argue that while it can mimic stylistic elements, AI lacks the depth and authenticity inherent in human creativity. Kathryn Bromwich, a writer and editor at The Guardian, said Altman praising the new model is like “parents who pretend to be impressed by their children’s terrible art.”

Other experts have also raised concerns about potential copyright violations. In fact, OpenAI is already facing multiple lawsuits alleging unauthorized use of copyrighted content to train its models.

Reece Hayden, an analyst at ABI Research, believes OpenAI’s venture into creative fields is driven by its failure to monetize its current developments in numerical subjects. Last month, the company praised ChatGPT-4.5 for its improved emotional intelligence and creative thinking, as opposed to previous advancements focusing on accuracy and predictability.

Hayden is pessimistic about AI creative writing, saying “it’s likely to experience significant backlash from creative industries as their intellectual property concerns are seemingly coming true.”

Meanwhile, the UK government is actually considering policies that would allow AI companies like OpenAI to train on copyrighted works without explicit prior permission. OpenAI and Google have also recently asked the US government to allow them to train AI models on copyrighted material.