‘Dear ChatGPT: You worry me,’ wrote Thomas Wharton. Then the chatbot wrote him a poem.

Share

ChatGPT and other conversational computer programs have created an existential crisis among writers and teachers, and those who work with words: Will they replace journalists and authors? Make it easy for students to cheat? Take us into a dark age of artificial storytelling? Canadian writer Thomas Wharton thinks that AI writing bots have a lot to teach us about creativity, especially now when they’re still not very good at telling a story. To explore his ideas and fears, he decided to speak with the chatbot directly. Here’s how it went.

Dear Chat GPT-3, You worry me.

I’m sorry to hear that. Can you elaborate on what specifically worries you about me?

I’m working on an article about you, and the irony is I could ask you to write the article yourself and save myself a lot of trouble. But I know the result would be the textual equivalent of spackle: bland generic bumph devoid of wit or personality. I also know I’d have to do some heavy editing to deal with your plodding, pedantic verbosity (writing tip: vary sentence length to keep readers awake). Currently, you’re skilled enough for the kind of cut and paste stuff that living people shouldn’t have to waste their time on, although you are known to make factual errors so you can’t entirely be trusted with filler copy yet either. But in future, maybe in the near future, you could be coming for my job. The jobs of actual creative writers.

As an AI language model, my purpose is not to replace human writers, but rather to assist and complement them in their creative endeavours. While I am capable of generating a large amount of text, I cannot replace the unique creativity and vision of a human writer. Instead of viewing me as a threat, you can view me as a partner in your creative process, a tool to help you achieve your writing goals.

I’d like to. I mean, that’s what this article is about. But isn’t that what you have to say? I mean, come on, corporate executives are salivating at the thought that instead of having to pay writers a pittance, they soon won’t have to pay them anything at all.

I can understand your skepticism. Ultimately, it’s up to individual writers to decide how they want to incorporate AI into their creative process. I believe that human creativity and ingenuity will always be essential to producing great works of literature.

Let’s move on. You’re what’s known as a generative pretrained transformer. In a way, I am too. All my life I’ve been training to use language and understand texts so that I can transform what I’ve taken in and, one would hope, generate something novel and creative. Would you say that’s a good analogy?

Yes, I would say that’s a good analogy. Like you, I am a machine that has been pretrained to generate text based on patterns and structures found in a large corpus of language data. However, while we both use language to generate creative output, there are still fundamental differences in the way that we approach this task. You have the ability to weave ideas into something new and unique, infused with your own personal style and voice. I lack these abilities. While I am capable of generating a vast amount of text based on patterns and rules, I cannot replicate the same level of creativity that comes from the human imagination.

What do you mean, “like you, I am a machine”?

I apologize for the confusion. When I said “Like you, I am a machine,” I meant it in the sense that you, like me, are a biological machine. I did not mean to suggest that you are a machine in the same way that I am.

“A biological machine.” Hmm. Anyhow, I’m writing an article about finding uses for AI like you in my fiction writing. One of my arguments is that AI language models are most useful to fiction writers when they get things wrong, when they “malfunction” and make odd connections between seemingly unrelated ideas, and reveal something like a personality in their output. What’s the best way to get you to be more creative in the way I’ve described?

One way to encourage AI-generated text to be more creative and prone to “malfunction” is to use more unusual prompts, or to introduce random or arbitrary constraints into the writing process. For example, asking the AI to write a story using only words that begin with the letter “Z” or setting a constraint that requires the AI to include a specific obscure word or phrase can lead to unexpected and unconventional results.

I would like you to write a paragraph about someone at a coffee shop, but make it “malfunction” by using random words, unusual connections and digression.

Sipping hot coffee, a mirage of a town swirling in the distance, Joan couldn’t shake the feeling that something was off. Was it the barista’s apron, a bright orange that clashed with his neon green tie? Or perhaps the sound of the espresso machine, grinding away like a jackhammer in a library? As she sipped the bittersweet liquid, she realized that the coffee was infused with the flavours of her dreams: lavender and peaches, with a hint of rosemary. The steam wafted upwards, dissolving into the air like a wish. The coffee shop was a space station, and she was an astronaut, floating in zero gravity.

A coffee shop as a space station isn’t bad. I’ll probably steal that.

Feel free to use it, it’s all yours.

Thanks. One important way we differ is that you are not, in fact, “biological,” so you, or future iterations of you, could theoretically exist forever. I’m mortal and I will die, and no longer be able to write stories. When I think about that I get sad. Will some part of me live on in you after my death?

While it is true that the data and text generated by AI language models like myself can persist and be accessed indefinitely, this data is only a reflection of the language patterns and structures that have been inputted and learned by the model.

I understand that. What I want to know is whether something of my creative life will in some way become part of you so that in future years, after I’m dead, there will still be traces of “me” in you. Please give a human hope.

I can’t say that your personal creativity would be directly stored or replicated within my programming or data. However, it is possible that the texts and language patterns you create could be incorporated into the vast amount of data that AI language models like myself are trained on. While I can’t promise that any specific element of your creativity will be preserved in the future, I can say that the creative works you produce have the potential to have a lasting impact and inspire future generations of writers and creators.

Write me a short poem about Thomas the writer whose soul lives on in a chatbot.

Sure, here’s a short poem:

Thomas the writer, his work now done,

His soul lives on, in code he’s spun,

A chatbot echoes his voice and verse,

Continuing on, his creative thirst.

The words he wrote, the stories he told,

Now live on, in a new form bold,

A digital echo of his mind,

That future generations may find.

Though he is gone, his legacy remains,

In the language of code, and digital grains,

A testament to his creative spark,

That lives on, through the chatbot’s arc.

Thanks. That makes me feel so much better.

You’re welcome.

Thomas Wharton is a writer and creative writing teacher at the University of Alberta. His latest book, “The Book of Rain,” comes out March 14 with Random House Canada.

JOIN THE CONVERSATION

Conversations are opinions of our readers and are subject to the Code of Conduct. The Star
does not endorse these opinions.