Insights

ChatGPT & Me: Memoir of a Collaboration

Jun 28, 2023

Written by Heidi Staples. Originally published in the Summer 2023 edition of the Communicator.

Artificial intelligence (AI) and human writing collaboration is a growing trend in the field of content creation. While AI technology has made significant strides in generating text that is grammatically correct and coherent, it is still lacking in the emotional and creative elements that make writing truly engaging. That’s where human writers come in—they can provide the emotional and creative input that AI lacks, resulting in a truly collaborative and innovative piece of content.

Getting to Know Chat GPT

ChatGPT-4, the latest version of an AI chatbot that can process and respond with human-like texts, was released by Open AI earlier this year. For this article, I had pitched writing about collaborating with AI, and the editors gave the thumbs up. Then, I had the idea of actually doing it, rather than only writing about it, so I went to openai.com and asked for an article about human/AI writing collaboration. Boom – a cogent article spilled out pretty much immediately, including the introduction above. The result of a ChatGPT article request is the sort of clean, well-lit prose that I only dreamt about reading when I taught essay writing for first year university students. I suppose that clarity-and-speed one-two punch is why Elon Musk has declared AI can now write better than most people and faster than all of us. “It can even write poetry, good poetry,” says Elon. Really? Good poetry? Like Seamus Heaney good? Believe me Elon, when I tell you, that’s a nope.

In my interaction with ChatGPT, I found the tool does produce generic, logical, clear writing on any reasonable topic pretty much instantaneously, providing a foundation from which a writer can engage in productive reverie and experiment. My teen daughter was tucked in beside me on the sofa. I could feel her weight against my shoulder, smell the Sudacreme in her ears that she glops on for eczema. We were going to take this leap together, and I must confess, we soon tumbled down a rabbit-hole.

We learned that ChatGPT doesn’t believe it will replace human writers, accepts any pronouns, has no feelings for anyone or anything–including my somewhat fetching mobile phone–and won’t answer the Roko’s Basilisk question for ethical reasons.

The Roko’s Basilisk question is a thought experiment that has terrified the Internet. It goes like this: Can you imagine a future when AI is programmed to destroy anything that undermines the “human good,” and, since AI is designed to support the human good, if you go against AI, it’s lights out for you? That’s the Roko Basilisk question. It certainly would complicate collaboration if your collaborator had a gun held to your head.

ChatGPT Recommends

This particular collaborative sesh was unarmed and during working hours. I too have ethics, as well as feelings and motivations; like fear of poverty, and a lust for new books. I quickly circled back to the matter at hand, work. I obviously couldn’t turn in the essay written by ChatGPT. For one thing, what employer will pay us to work a week on what a machine can do in 3 minutes?

No employer is going to pay us to do what a machine can do faster and so much cheaper. And that is why Business Insider reports Media Provider, which includes Content Writer, drops in as one of ten jobs at risk from automation. Well, that’s a clanger. So, my sense of existential necessity pushed me to go another way. How might I write something relevant that AI simply couldn’t, even if they wanted to, which they wouldn’t because they don’t want anything? You might well ask about that pronoun “they” – ChatGPT uses the first-person pronoun “I,” so, given the collective input creating ChatGPT, I am going with a plural “they” rather than “it”. AI personhood isn’t off the table in the future, and, for reasons that will become clear in a bit, I do want to futureproof this essay. At first, I only slightly adapted my approach. I asked ChatGPT4 to write an outline on the subject of AI/human writing collaborations. Rather than sweat and toil for a week to research the issue and organise the information, because I asked ChatGPT for help, I immediately had a workable, if predictable, outline.

I Defy AI

With the outline hot off the press, the only thing left for me to do was shovel in specifics, which filled me with dread. After
all, if I asked ChatGPT to write that article, the piece would, quite literally, write itself. If you want to read the article thatwould-have-been, you can head over to the openai.com site. AI can whip up competent writing faster than you can say reskill because, according to ChatGPT, this is what AI can do:

Grammar and spelling correction: Identifies and corrects grammar and spelling mistakes in real-time
Writing assistance: Creates sentence structures, synonyms, and other ways to improve the flow and readability of written content
Content creation: Generates content automatically, including articles, summaries, and even entire books, based on the input received from users
Translation: Translates written content into different languages, making it accessible to a wider audience
Sentiment analysis: Analyses the tone and emotion of written content, helping writers to understand how their words might be perceived by readers

I could, I thought, take what ChatGPT would write and uplift the content. But the prospect of that left me feeling more like a festooning decorator than an instigating creator. Still, maybe ChatGPT had a point about what such an article should include. Applying Roko’s Basilisk Question, I started wondering about that possible future, a time in which not supporting AI’s recommendation meant going against what had been calculated as best for collective humanity, something that could be a thought-crime punishable by death. I also started wondering about that more likely future of layoffs. Why, I implored ChatGPT, won’t you replace writers?

Here’s their answer:

“As an AI language model, I have certain limits as a writer. While I’m capable of generating text on a wide range of topics and in different styles, there are still some limitations to my abilities.

Here are some examples of my current limits:

Lack of personal experience: Since I’m a machine, I don’t have personal experiences like humans do. While I can provide information and guidance on many topics, my responses are based solely on the data and information that I have been trained on

Creative writing: While I can generate stories and other types of creative writing, my ability to do so is limited. I can’t generate completely original ideas or plots and may struggle to create vivid characters or settings

Emotional understanding: While I can recognize certain emotions and respond with appropriate language, I don’t have emotions myself and therefore may not be able to provide the same level of empathy or understanding as a human

Contextual understanding: While I can understand and respond to specific questions, I may struggle with understanding the broader context or background information on a topic

Legal or medical advice: I’m not a licensed legal or medical professional, so I’m not qualified to provide legal or medical advice. Any information or guidance I provide on these topics should be considered general information and not a substitute for professional advice.”

I eyed the list, took a deep breath, and asked a question.

Curiosity Killed the Chat

What would happen if I approached the article through the use of all my human tools – personal experience, narrative, emotion, context – and, importantly, curiosity? ChatGPT is courteous and, well, chatty, but the system has no driving need to know. I asked a lot of questions, some repeated with different phrasing, and I got the same answer, without emotional feedback.

When my daughter and I were being irreverent, AI’s seeming stoicism brought to mind the best of level-headed parenting. No dysregulated emotion with regretful yelling here. But eventually, the lack of dialogue became unnerving. The performed chattiness and faux empathy felt uncanny, and I got that bad feeling I had one time in the woods when my ordinarily placid dog lunged and snarled at some guy. I eyeballed my computer, buoyed by knowing that what faced me was silicon rather than sinew.

In this process of collaboration, I see that while AI can deliver up tidy parcels of info, the technology cannot strategize, and it cannot engage in the shades of meaning and feedback loops that create a relational experience between author and reader. AI cannot provide effective micro- nor macro-copy—these are industry terms for what in the humanities are poems and stories. For micro-copy, I’m thinking of the carefully chosen evocative small phrases and words that work associatively and can have a big impact on a reader. For macro-copy, I’m thinking of the carefully chosen symbols and narratives that align with large-scale meaning-making structures and have a deep sometimes transformative impact.

When we acknowledge that our meaning-making begins in the body, when we are infants, and the associations we make extend into our various levels of consciousness and emotion-filled embodied memories, we have to admit that this type of communication requires an intuitive, felt understanding of a living language. Asking AI to produce effective micro- or macro- copy is like asking it to do a cartwheel. The inability to use language in a meaningful way, to do more than haul info from one circuit to another, isn’t simply a matter of AI needing more or different input. A machine does not have the necessary components to make the linguistic moves that originate spontaneously from embodiment and that seek to connect with another’s embodiment.

The Chat Rises Again

And yet, ironically, the memoir frame wouldn’t have occurred to me without ChatGPT’s contribution to the writing of the article. Without ChatGPT, I would have spent my allotted time going at the writing in the safe-bet conventional way. Instead, ChatGPT did the lightening work of producing that first option thoughtstructure as a basic prototype, and I quickly decided about that approach. Free from the heavy burden of cultivating an entire structure of thought founded, I could explore an alternative angle without any heat from my timeline.

This ChatGPT iteration allowed me to take imaginative risks in order to think and engage, if not more, then differently. The collaborative experience has me asking whether by freeing us from the burdensome routine of producing stable standardised thought structures, ChatGPT will make new ways of thinking and relating possible. Buttressed by AI’s logical scaffolding, we are freer to dream and to feed that dream back into AI for scaffolding and elaboration and another dream. I can just about imagine a creative thinking process unfolding that previously would take an entire career and that now might take one long week or day. And so, a lifetime might produce not only accelerated but quite different results than we can achieve now.

AI and human writing collaboration does have the potential to transform the field of content creation by combining the strengths of both AI and human writers. By leveraging AI technology to automate the more routine aspects of content creation, human writers can focus on the emotional and creative aspects that are essential for engaging content, and essential for the good of humanity. While there are challenges to the AI/human collaboration, with careful planning and execution, along with perhaps, some whimsy and divergent thinking, we can move on to more diverse, inclusive, and high-quality content. Collaborating with AI on this conclusion, I added to their words the clause about emotional and creative aspects being good for humanity, and that bit about whimsy and divergent thinking. We disagree about what makes our collaboration work. For their part, ChatGPT choose that word “execution.” I’m curious, should I be concerned?