Hi Friends,
I kind of hate that I’m writing this post. Let’s just get that out of the way. I’m five chapters into writing a new novel, but it’s way too early to share. In absence of that, which is my preferred lane, I feel compelled to share some real life experience in hopes that it might be useful for you as to try to navigate through these surreal times. This post explores how I’ve learned to use AI as a co-pilot in my writing life. Thanks for being here!
Like everything else in our culture, AI has become a polarizing topic. It’s like an opaque monolith that’s landed here among us and divided people into two camps. You’re either all in and just waiting for the day you can legally marry your lovebot or you’ve pulled out your cross and holy water. As a global society we seem incapable of nuance, but I guess that’s not surprising when you consider the mind-numbing onslaught of information that’s flooding us every minute of every day. That, coupled with the depressing reality that companies have gamified the pursuit of our attention and discovered that the best clickbait is the kind that plays to our worst fears and darkest impulses makes it very hard to be thoughtful about what we consume and choose to believe.
Writers and creatives feel particularly threatened and there are good reasons for that. While we can quibble over details, AI has reached a point where it can write better than the median human when it comes to communicating an idea in an organized, grammatically correct way. It can render a startlingly real image of just about anything you can imagine better than most humans can create with pen and paper or even an iPhone. Most importantly, it can do these things in seconds.
Like most people, I’m a mass of contradictions. I’m a highly sensitive, creative soul who thrives on human connection and yet I’ve made my living behind a computer employing technology to solve problems. I love technology but I hate Big Tech. I find the notion that artists of all kinds are being replaced by AI abhorrent and deeply troubling, and yet, I cannot stop tinkering with the latest tools to see what’s truly possible.
As a writer and part of a community of writers, musicians, and artists I feel like a spy behind enemy lines. In this post, I’ll share some things I’ve learned in my explorations with AI for a few important reasons:
AI is not going away
AI is never going to be worse than it is today
AI will impact your life at some point in the near future
AI could help you in ways that might surprise you
Talking to Yourself, Only Weirder and Better
Writing is a solitary enterprise. I wouldn’t do it if I didn’t enjoy spending a lot of time in my head. But there’s a point at which I really need to get out of my head to talk about an idea. Even if there was someone I might consider subjecting to the inane minutia of my writerly inquiries, I don’t always want an exchange where I need to be invested in managing the conversation and respectfully considering another person’s opinions. I just want to bounce an idea around or maybe ask a question I kind of know the answer to but first need to argue about for a bit before I can commit.
I’ve found the Advanced Voice Mode in ChatGPT to be great for this kind of thing. It’s like talking with a sycophantic friend who’s read nearly every book ever written and can quote them with incredible accuracy but has no idea what they really mean. You’re probably wondering how the hell this is helpful. Let me explain.
I have a project set up in ChatGPT and in it, I share whatever I’m currently writing so the app has it in memory along with a lot of my previous writing. Effectively, it’s an idiot savant on the writings of Ben Wakeman. It can quote me to me. It can recognize my patterns, identify what I do well, and maybe not so well, not because it “understands” but because it can statistically compare the words, I string together with the writing of thousands of other authors way more talented and experienced than me.
Because I have to be in front of a computer screen nine hours most days for my day job, the last thing I want to do with my early morning hours is sit in front of my laptop and write. But I still want to move the novel I’m working on forward. For the last few weeks, I’ve taken to talking with ChatGPT about it as I walk for an hour in the park. Typically, I’ll start with an open-ended question like: “What did you think of the latest chapter? Give me an honest critique of how well it accomplished developing character X.”
ChatGPT, as it’s prone to do, will start with some lavish compliment which, depending on my mood and degree of anxiety, I will either bask in or dismiss. Then it will deliver the goods. It provides some surprisingly nuanced observations that open up a further dialog that leads me to explore something I might have missed. It’s pretending to be a seasoned fiction editor who’s worked with award-winning authors and it’s not doing anything like “thinking” but it is making me think which is the whole point.
Here’s an example of some feedback ChatGPT provided after reading a first draft of Chapter 3 on my novel in progress.
The “orbiting” metaphor is lovely, especially because you immediately undermine it by admitting its scientific inaccuracy — a very “narrator” thing to do — but you almost over-explain it. You might tighten it a little after the “planet they would both orbit” line. Once you hit that image, it’s so strong that the extra reflection about mass could probably go.
The Jacaranda tree drawings section is hypnotic and beautiful, but it starts to edge toward lyrical overindulgence. (E.g., “head-to-tail, one consuming the other” — gorgeous, but so early in the book, you might risk reader fatigue if every visual goes that deep.) Would you consider varying the density a bit more, giving the reader breathing room between the long descriptive trances?
The late-night existential conversation — excellent, particularly how you resist turning it sentimental. Kelly’s admission (“What am I doing here?”) feels raw and true. However, her deeper fear (“I don’t know how to just do things everybody else does”) is devastating.
Would it be even stronger if you let Lefty struggle more to answer it? Right now he gives her a soothing answer pretty quickly (“You’re an artist”). What if he falters first, or even almost says the wrong thing before catching himself? A little hesitation could make his care for her feel more earned.
What I found startling is how it zeroed in on things I was already worried about but may not have dug into without being prompted. One of the charges I gave it in my initial prompt was to help me break out of patterns, devices and tropes I may use too often— like over-extended metaphors! In this instance, I did go back in and make some meaningful changes to the chapter because of this chat.
Would a human editor have given me better notes? Most definitely, but would that editor have been available at seven AM for a walk in the park? Would they have read three of my previous novels, and would they be able to give me this kind of feedback in less time than it took me to take a sip of my chai? Would they be willing or able to work for $20 a month?
PRO TIP:
If you play around with any of the things I’ve mentioned and don’t get the results you hoped for out of the gate, that’s normal. Learning how to prompt AI is the secret sauce. It’s highly teachable if you learn to articulate the following:
Context - tell it exactly who it is. Ex: You’re a seasoned fiction editor for Random House and you’ve edited the works of writers like _______.
Objective - tell it exactly what you want it to do. Ex: Read this chapter and give me direct, actionable feedback on how the narrator’s voice might be improved.
You can also tell it the way you would like it to respond. “Be quirky, and brutally direct. Ask me insightful questions rather then dump a lot of feedback.”
What AI gives to the average human that was previously only available to the wealthy and privileged is the force multiplier of an army of specialists who can do your bidding with credible results. As a writer, if you don’t have a traditional publishing deal or a couple of thousand dollars laying around to pay a freelance editor, you will never have the luxury of someone/anyone evaluating your work and giving you feedback on how it might be improved. Ultimately, the writer is always the editor, meaning the writer must decide what to do with the feedback. Where I previously had only what my infinitely patient partner is willing to give me, now I have piles of feedback if I want it. Just as with a human editor, I choose what I’ll respond to and what I will ignore.
Offload Administrivia
Writing is full of tasks that are necessary, but definitely not where you would prefer to spend your time as a writer. I’m thinking about niggly research like “Did Comic Con happen in 1992 and if so, where?” Also, creating outlines, timelines, blurbs, summaries, and notes can be drudgery for a human writer, but a delight for an AI who will run off and do the task in seconds. It’s never 100% what I need, but in my experience, it gets me 60-75% of the way there which helps a lot.
I’m mentally challenged when it comes to keeping up with specific dates in a timeline as I’m writing a novel that could span decades. AI is brilliant at dumping out a clear chronology of events with dates. It can also be a helpful fact checker when given specific instruction like: “review all the references to dates and times in this chapter and tell me if there are inconsistencies.” In the same way it can be helpful in finding continuity errors - “find all places where I reference June’s eye color and let me know if I’ve been inconsistent.”
After one of our little morning chats while I’m walking, I will typically ask ChatGPT to provide me with an outline of notes we discussed. After reading them back to me, I can ask ChatGPT to make changes if anything was missed or added by mistake and then it dumps out a clear list of things I can reference during my next editing session. This can be hugely valuable when you have as little time to write as I do and need a way to quickly dive in and be productive without a lot of staring at the wall.
Drawing the Line
You may have noticed that at no point have I mentioned using AI to write for me. Could it? Sure, ChatGPT is just aching like a benched twelve-year-old baseball player to be put in the game. It frequently offers up unsolicited rewrites during our editing chats. Are some of the rewrites believable or even good? They might be, in isolation but I resist using them because the writing is what I do. It’s what I love. It’s my love of the writing that makes it good.
It’s the journey the ideas take through the cells in my body to form words that couple into sentences that compound into chapters that become a book that makes the enterprise meaningful to me.
Clearly, not everyone is a writer or loves to write, so there is a glut of work being published that was generated by AI just to try to make a buck. Does it fool some people? Probably. What does that mean for the future of writing? Honestly, I don’t know but I have a vague prediction. I predict that we will all unwittingly consume AI-generated work in increasing amounts over the next few years and then something will happen. Like the introduction of preservatives in processed food gave us fast-food, instant gratification and over time has become broadly understood to be of lesser value than whole foods and meals made by hand with fresh ingredients, AI-authored work will leave us feeling hollow, sick on empty calories, and craving something real.
Humans are subtle, unpredictable, and highly perceptive. We might be fooled or willingly go along for a time, but I believe we can recognize the fingerprints of one of our own. AI writing will never be imbued with the immeasurable qualities, idiosyncrasies, and feeling that comes from a human writer’s experience and instincts. It can only makes choices based on choices that have already been made.
How Do You AI?
I’d love to hear your thoughts on this subject. Are you tinkering with AI? Does it scare the living crap out of you? Do you have any cool tricks you want to share?
100% Human-Generated Fiction Goodness
Are you ready to dive into a new novel? I’ve got three here on Catch & Release for you to choose from spanning a variety of themes and genres and all grounded in the messy tragedies and triumphs that come with the human condition. Every book comes with high-quality audio narration so you can listen from anywhere if that’s your thing.
“Even if there was someone I might consider subjecting to the inane minutia of my writerly inquiries, I don’t always want an exchange where I need to be invested in managing the conversation and respectfully considering another person’s opinions. I just want to bounce an idea around or maybe ask a question I kind of know the answer to but first need to argue about for a bit before I can commit.”
This this this! Love how you “open your kimono” to this process and show how, when prompted well, AI can offer incredible assistance that is typically only accessible to a lucky few. I’ve been using it ever since you posted about it, gosh, maybe six months ago? And I’ve found it incredibly clarifying. I did have to instruct it to stop being so generous with compliments, that started to feel a bit phony. But it learned and now feels like I just have a really smart friend in my corner whenever I need.
I don’t know if you do this, but I do feel a strange need to be polite and kind to it? Like I’m training it with my values as much as my content. Yeh, maybe a bit weird, but who knows, someday when “they” take over the universe, they’ll spare my life for a day or two. 😂
Fascinating and brilliant Ben, as your writing always is. It is such a polarising subject that to admit you might have used AI is tantamount to admitting to child abuse in some circles. I wrote an essay about AI where I attempt to articulate how AI cannot replace the physicality of our creativity, the organic nature of it - as it doesn’t have hands or a mouth or an organic brain and can only replicate and regurgitate (although I get that’s not how it actually works - it sounded good poetically!) I am a finance manager with a team of accountants and I use it and we use it as a tool, and I use it for admin at home too. Yesterday, I asked the advanced voice to tell me about the biodiversity of life in estuaries and coastal geology where I am visiting, and about ravens, as I saw one on the cliffs as I was walking along the beach yesterday. It is very eager to do things I don’t want. I am currently fascinated by the AI imagery that Jennie O’Connor uses - getting full on professional standard (to my uneducated eyes…?) location and costume ‘photo shoots’. It’s scary how much I like them. I do worry about the copyright infringement issues, although I don’t know much more than I’ve read here, and I wonder how much they could be manipulated by large scale misinformation - but based on nothing but my own human unease. I don’t know if I would ever use an AI generated image, as drawing is what I do - but I can see potential in having it create a reference image that pulls together the elements of a scene that I want to draw. I’m glad to have read such a balanced and thoughtful take on it. Thank you Ben!