Three unexpected ways chatbots can spark creativity

Sometimes, deadline pressure can get you stuck—the harder you try to come up with an idea, the deeper the rut. But what if you tried stepping off the path of linear thinking?

Humans can get bogged down by a desire to make perfect sense. Chatbots, not so much. They can let go of logic on a dime. “The problem is that users expect AI chatbots to be both factual and fluid,” writes Lauren Leffer in Scientific American. 

So, as an experiment, let’s shift our expectations. Instead of looking to chatbots for factual accuracy, let’s look at their hallucinations—those eerie moments when chatbots respond with random nonsense—as food for thought. Here are three ways to reframe those freakouts as the first step in your creative process.

Let the weirdness steer you

Last year, author Stephen Marche made headlines by using AI to prompt the mystery novella Death of an Author. To develop the story, Marche relied on a combination of ChatGPT, Cohere, and Sudowrite. First, he’d instruct one of the chatbots to write in various styles, then he’d switch to another to rephrase the original output. As he kept iterating, he leaned into the unknown and welcomed unexpected twists. 

“There were times where the machine was creating something that I would never have created—and those are the really interesting moments,” he says. “I don't mean to sound weird about it—but it was like I was getting an alien to write a short story.”

“At one point, I got it say, ‘She walked down the hallway like an LP going back into a sleeve,’ which is a great line,” he says. “But I'd be doing a scene of a woman going on a dock somewhere and it would give me a description like, ‘She's wearing a yellow dress.’ It was very specific, like it had an idea. Like, ‘Let's make this character this way.’”

Marche notes that although it feels like the AI has intentionality, the truth is that even its developers don't completely understand its workings a hundred percent of the time. 

“It does not participate in the scientific method,” he explains. “It's not something that is understandable. So what comes out of it is kind of magical. I don't mean that in some metaphorical way. I think it's actually a force that we don't really understand, which is the name that we give to magic.”

Make a map of where you don’t want to go

Some creators find AI useful as a kind of mental GPS. In the same way that  Google Maps can show you different route options, chatbot collaborations might not provide the fastest way to get where you’re going. But sometimes, the “scenic route” is what you really want.

For example, Jordan Harrod is a PhD candidate at Harvard-MIT who creates educational videos that help viewers understand the impact of AI and algorithms. On top of using AI to edit videos, manage her schedule, and plan trips, Harrod uses chatbots when she’s working on a draft of a script or an outline of a talk, but she’s not sure what she wants to say. 

“What often happens is that whatever is generated, I look at and I'm like, this is not what I want,” she says. “But now that I know what I don't want, I can start writing things that I do.”

She also relies on chatbots to augment her creativity when she’s struggling with writer's block. “If I'm running into a wall with the ideation process, I’m probably going to have Notion AI do it—but I might also use ChatGPT.” That’s why, when she’s looking for a starting point, the chatbots often help her solve the blank page problem. They can give her a first draft that sparks a reaction to get the wheels of imagination turning. 

Make the nonsense useful  

Ever had a creative breakthrough happen by mistake—like when you're playing guitar or piano and your hand lands in the wrong place? Sometimes, it sounds more interesting than what you’d planned. And for creatives, chatbots can deliver an endless supply of those happy accidents. 

“I've certainly changed my path from what I was looking for because it weirdly misunderstood what I was saying,” says Kevin Kelly, co-chair of the Long Now Foundation and author of 14 books, including What Technology Wants. “Language is very ambiguous and word order makes a difference. So right now, happy accidents are part of the norm. It's something you should expect.”

Even before ChatGPT, Kelly was using AI tools like Sudowrite to create first drafts from a prompt.

“Most of the time, it's surreal nonsense,” he says. “Then you have to make it useful—bend it or add a little more. So there's a conversation. They’re good at filling in details that you wouldn't even have thought of in a small, constrained way. It doesn't have enough attention to continue something over multiple scenes. So that's where the human is involved. You're kind of directing: Okay, now we go here. It's a co-creation.”