OpenAI is working on a new AI model Sam Altman says is ‘good at creative writing’ but to me it reads like a 15-year-old’s journal

Ever since I was a wee sapling, I wanted to be a writer. The goal was originally to become a published fiction author, but then I discovered videogames, and that’s hauled me down a merry little rabbit hole that I’m not in a particular rush to exit–truly, a time honoured tradition for creatives everywhere. Well, OpenAI’s latest project is angling to automate that dream.

Company CEO Sam Altman has revealed that OpenAI is currently training a new Large Language Model AI geared towards creative writing. On X, Altman shared his initial prompt and the AI’s resulting paragraphs of purple prose which he described as “good” (via TechCrunch).

There’s currently no solid timeline for a wider release of this model, though Altman sounded optimistic of the AI’s capabilities. With regards to the lengthy progress sample he shared on X, he wrote, “This is the first time I have been really struck by something written by AI.”

With the knowledge that OpenAI has in recent months been moving away from its initial nonprofit mission, (especially in light of the billions of capital the company reportedly burned through last year), it’s hardly surprising it’s seeking to diversify its portfolio. OpenAI is perhaps best known for ChatGPT, their AI assistant Operator, and an ongoing saga involving Elon Musk. While its already publicly available AI products are meant to sound at least sort of conversational, branching out into creative writing models is a new one for the company–though I personally don’t share the CEO’s optimism.

Why? Well, I could make some cheap shots like pointing out how the AI amateurishly deploys a ‘rule of three’ list twice in the first paragraph alone. Or I could complain about how the metafictional framing is weak at best, and manipulative at worst. Or I could just dismiss the whole thing as massively overwritten, bringing my own less than stellar rendition of the ‘rule of three’ to a close.

…I could also make some allusion to the novel writing machines mentioned in George Orwell’s 1984, but I think that comparison is a little overplayed, don’t you? Instead, here’s my two cents: as a regular, ordinary human who has written about my own lived experience with grief, the whole exercise feels pretty gross. Why grapple with such tricky emotions when you could outsource it to an AI model? Why connect to the words of another person when an AI can show you a funhouse mirror? No thanks!

Bottom line, any truly in-depth criticism of this LLM sample feels a bit like I’m already giving the AI far too much credit. Besides, it’s hard to take the response sample at face value when there’s no real telling how much human editing has taken place before I’ve clapped eyes upon it. Who’s afraid of AI’s purple prose? Not me.