(This is part of my series on AI. Click here to read about AI in general and Google AI Audiobook Narration. Or, sign up here to get these delivered directly to your inbox!)
ChatGPT is the talk of the century! Well, maybe not the century, but it’s definitely the talk of the year. ChatGPT has been available to regular people in some form since June 2020, and was built by a company called OpenAI. Their language model is one of the most sophisticated in existence, and they have continued to improve it since their launch. Currently, ChatGPT4 can pass the bar exam.
Mostly, ChatGPT can answer questions, draft emails and blog posts, or summarize large concepts. And depending upon the parameters, it can do so in a wide variety of voices and styles.
Take this meme, for example, in which ChatGPT has been instructed to respond to the user’s requests… as a pirate.
ChatGPT is not the only text-based AI. For example, NovelAI and Sudowrite are designed to aid in drafting longer form fiction. You may have also heard of Claude, which is in beta, but is more like a combination of ChatGPT and these longer form tools. GoogleBard is an alternative to ChatGPT but built on a different model, and so has a very different vibe. Microsoft also has Bing Chat (which provides citations!), and there are other lesser known tools like Jasper.ai, ChatSonic, and YouChat.
These tools are simple to use. All you do is type in your prompt and they respond. If you don’t like the response, you change your prompt.
The key to successful text generative AI is in the prompt. You have to be able to communicate with the tool exactly what you’re looking for in a way that will produce the desired outcome. This requires some practice, but it is a skill you can develop.
The biggest objection among authors is that these text-based tools are “doing the work for them.” Many writers use this reasoning to suggest that the tools should never be used in any way.
But let me offer a comparison: ghostwriters.
Many writers (James Patterson, for example) use ghostwriters to help them produce work faster. In his MasterClass, James Patterson goes over his process in depth, which basically involves writing an outline, passing it off to a ghostwriter who writes it in sections, and then revising the written manuscript one section at a time.
There are some writers who are using AI tools in the same way: they come up with the idea and the structure for the story, then use the tools to spit out the rough copy, which they then revise and edit to craft it into something that matches their vision for the story.
But, there are other writers who are using it for research, idea generation, or to help them move through tough sections, but still do all the drafting themselves—just like how most of us don’t use ghostwriters, but we do use Google Search, libraries, and friends. These writers treat the AI tools more like a personal assistant or a writing buddy.
Are some writers simply copy-paste-publishing? Sure. But some writers are also writing 50k words during NaNoWriMo and publishing that without editing and revising. There will always be people who try to game the system, but that doesn’t mean it’s the end of the world as we know it.
Back when indie publishing first became a thing, there was a surge of writers publishing without sufficient editing. I can honestly say, I was one of those writers. The process of publishing a book in this method was new, I lacked funds to hire professionals to improve my book, and so I just went for it. I have since gone back, cleaned up the copy, done professional covers, and improved my process for future books.
Since that era in publishing, the indie author community has changed direction. There are courses on how to publish books that are professional and high quality. Due to the Big 6—Big 5—Big 4? buying each other up, there is a glut of quality designers, editors, and experts in publishing available for hire by indie authors. We have developed processes and procedures, and the surge of low-quality books that were originally being published has decreased significantly.
The same will happen with AI-generated books. The good ones, the ones that feel like art, will rise to the top, and the rest will sputter and die. Processes and procedures will be developed. Authors will figure out the best ways to use these tools to enhance their own process and create quality stories that readers love.
We just have to be patient.
My personal use of these tools tends to focus around reducing distractions and cutting down on research. Instead of spending two hours digging through baby name sites, I instead ask ChatGPT for a list of names in a specific category or with a specific meaning, and just pick one. Instead of reading a dozen articles on knot tying, I ask ChatGPT “What kind of knot would you use for this purpose in this context?” Instead of getting lost following rabbit trails on Wikipedia, I ask ChatGPT, “What are the top philosophical thought experiments to help understand existentialism?” so I can focus my efforts on the most relevant concepts.
So far today, I have asked ChatGPT how old it is, how it would define “text generative AI” and what I should do to celebrate my birthday when it rolls around.
I have yet to copy and paste even a single phrase into a book, but my writing time has gotten about 30% more productive.
It’s important to remember that these language models aren’t thinking.
The idea that they can “think” is entirely science fiction—combined with a human tendency to anthropomorphize anything that feels human-like. Rather, they are doing math. After all, that’s what computers are good at. Calculating.
Because it’s text-based, meaning we read it inside our own heads; and because the AI is so sophisticated in its calculations and algorithms; and because of the immense quantity of data it was trained on, the tool’s output can often make it feel to us like it’s thinking.
But it isn’t.
It is simply running a predictive calculation based on the prompt you inputted, guessing at what response you are most likely to want from it.
Ultimately, one thing remains true: the human controlling the process has to make a choice. Just like in every other form of art.
And that choice is yours.
You choose what tools to use. You choose what prompts to input. You choose what outputs most reflect what you’re looking for. And you choose how you want to use those outputs in the final product—if at all.
Whether we are talking about text-generative AI, image-generative AI, audio-generative AI, or anything else—the fact remains that you must still make the choice.
Despite the incredible advancements of these technologies, they still require a human element. And that human element is what gives art soul.