As I sat down at my computer, I couldn’t help but feel a little nervous. I had heard so much about ChatGPT, the AI language model, and I was eager to try it out for myself. But I wasn’t quite sure what to expect. Would ChatGPT understand me? Would it be able to carry on a conversation?
I took a deep breath and typed in my first message: “Hello ChatGPT, how are you?”
I did not write these two paragraphs. ChatGPT did when I asked it to write an essay from the perspective of someone writing an essay about ChatGPT.
In its own words, ChatGPT is “an artificial intelligence trained to assist users with a variety of tasks and answer questions to the best of [its] ability.” Its most interesting feature? ChatGPT can write about anything. The possibilities are endless. Ask for an abstract of postmodern writing and the chatbot will spit out paragraphs saturated with skepticism towards metanarratives and the decentering of the human subject. Request a classic BuzzFeed-style essay about privilege in America and you get a listicle. Demand that it produce a screenplay about two Yale students fighting to the death with condiments, and in a scene, they will “clash in a frenzied dance of violence” in the Benjamin Franklin College dining hall, “jars of mayonnaise flying through the air as they fight for dominance.”
With access to nearly every web page created before 2021, ChatGPT is able to reference a vast array of texts for its automatic prose generation. It paves the way for hyper-efficiency enabled by machine learning, making various writing-related business processes a thousand times more seamless. In other words, it’s a technoliberal’s wet dream coming to life.
To others, however, this premise sounds less like an erotic fantasy and more like a dystopian nightmare. Much of the way modern life is organized will fundamentally change when the ability to write fluently is no longer a human activity, but instead a task delegated to computers. Since ChatGPT’s release, headlines have declared “Artificial Intelligence can write as well as humans.” What has this discourse made abundantly clear? In the face of new technology, the writer is rendered functionally obsolete.
But ChatGPT’s writing is about as efficient as it is stale and boring. When I asked for a creative essay on the feeling of loneliness on Christmas Eve, it churned out: “I feel nothing but loneliness and emptiness.” ChatGPT forgets the first rule in writing: show, don’t tell. As you may have noticed from the two paragraphs at the start, it writes entirely in simple statements. There is no unique voice or panache or flair, and the level of vocabulary is comparable to that of a middle schooler. Its structure, style, and content are formulaic. And even with an all-encompassing body of English texts at its fingertips, when I ask ChatGPT to replicate an essay by one of my favorite writers, Joan Didion, the program fails to mimic her distinct verboseness and stubborn preference for conjunctions over commas.
But will there be a future when this technology improves? Might the machine one day write as creatively as we do? Or perhaps—more frighteningly—could the machine soon steal the unique voice with which each and every one of us writes? According to the New York Times, automated tools can already identify the author of text with enough training data. A future where a writer’s style is not only detected but successfully mimicked does not appear too distant.
So, how do writers resist a system that threatens the unique ingenuity of human-written prose?
Make your writing as irreplicable as possible. Use more big words. More uncommon words. Aid your writing with synonym-searching arsenals like WordHippo or Thesaurus.com and give these algorithms more text to tokenize, more data to disassemble. Honestly, fuck all meaning: use words incorrectly. Create new ones.
Write MORE. And LONGER. Stop being concise. Make endless lines of prose. Delay the terrorizing computational powers of natural language processing. Weave into your writing metaphors that are tied strictly to their context, held tight in fleeting lines of text, so that they can never be processed as a discrete data point. Use the comma, the colon: the em-dash—and use them in innovative ways that are entirely novel—and wrong. Sabotage all the rules these algorithms follow.
Our machines have evolved. Our writing must do the same.