An AI Might Have Written This

As a writer collective, we’ve had AI on the brain—from my last piece on AI companion bots to Evan’s excellent essay on the AI value chain to Nathan’s exploration of the infinite AI article. 

Every has also been building Lex, a word processor with AI baked in. I started working on this piece before we launched Lex, but testing out this tool (among others) has shaped my perspective on the role of AI writing assistants for creatives. Try it for yourself: watch the demo and sign-up to join the waitlist (Every’s paid subscribers have priority access, so subscribe to skip the line).

In 2016, filmmaker Oscar Sharp and AI researcher Ross Goodwin created an experimental short sci-fi film written entirely by a neural network. Sunspring, the futuristic space comedy generated by the machine, was the result of an AI trained on hundreds of sci-fi TV and movie scripts. Its viewers noted that the actors in the film did a phenomenal job spouting gibberish, English words strung together with little meaning.

A year later, author Stephen Marche worked with two researchers to help him create the perfect science fiction short story, using an algorithm they had developed. It was trained on 50 short stories Marche himself wished he had written. Assembled by algorithm, the story fell apart under scrutiny from human reviewers who noted it was “full of unnecessary detail” and “sentences that don’t actually hold up when you read them carefully.”

More than six years later, in the midst of an AI boom, artificial intelligence has yet to invalidate human creativity. Writing generated by AI alone remains obvious or “off,” even with the development of Open AI’s novel GPT-3. Examples of creative fiction generated with GPT-3 are undoubtedly impressive, but even Open AI’s CEO has argued the response to the technology is overhyped. AI writing tools can’t yet produce cohesive scripts or books that replace the best authors and screenplay writers.