All of a sudden, everybody is talking about AI. Everybody has an opinion. They range from the deeply knowledgeable to the hilariously uninformed, from “the sky is falling” to “I, for one, wish to welcome and embrace our robot overlords.”
A lot of writers and other creatives are worried about it. AI seems to be churning out the sort of content we get paid for, at no cost to the company who owns the software. Some say it has the potential to eliminate how many creatives eat food and sleep indoors. There’s a sense of panic.
So let’s talk about AI. As of today (December 26, 2023), here’s the state of things as we go into a new year.
What AI is. What AI isn’t. What’s happening with AI, what that means for authors, and how this new tool can help us in our work. We’ll end with some predictions about what’s coming next.
Ready? Good. Let’s get to it.
What AI Is
When we watch (or write) science fiction, we think of AI as computers who have intelligence, autonomy, and free will much like a human brain — only faster, and in a more durable body. When we hear about AI as it’s being used now in the creative sphere, it’s tempting to think of the same thing.
But that’s not what’s happening. Here’s what is happening. Imagine you have an infinite set of Scrabble tiles. You pull out a letter S. Next you pull out a letter X. There are no words in English that start with SX, so you throw the X away. Next you pull out an H, which works so you keep it. O comes next. SHO starts many words, so you keep that. The next tiles are Q and Z, neither of which work. You discard them, then draw an E. You’ve made the word Shoe.
If you do that enough times, as the saying goes, eventually you’ll write Hamlet. AI does that, only at extreme speeds and with large databases against which to compare not just letters, but words, sentences, paragraphs, and structures. But no matter how fast it goes (and it goes very, very fast), it’s not doing anything besides that.
It’s not creating. It’s not thinking. It’s not strategizing. It’s just an incredibly fast monkey in a room with a typewriter, and the Library of Congress as a crib sheet.
What AI Isn’t
Okay. So AI isn’t the HAL 9000. It’s not Skynet. It’s not Number Five from Short Circuit.
It’s not even very accurate. Back in 2023, a lawyer got in big trouble because he used ChatGPT to prepare a filing for a case. ChatGPT wrote a brief that looked good on the surface, but referenced multiple cases which did not exist.
You’ll find similar factual errors in most AI-generated web content, which is commonly referred to as hallucinations.
You can try this yourself by asking an AI bot to write a 1000-word blog post about a topic you’re an expert on. Chances are you will find at least one important factual error, and several detail errors.
At best, AI turns out content like we had in the late 2000s when content mills were hiring people to crank out 500-word articles to keyword specifications to spoof Google. They were low-quality. They were written by people with no baseline expertise. They were everywhere.
Bottom line: don’t worry too much about AI taking our jobs this decade.
What Is Happening With AI
So, enough about what AI isn’t. Here’s what it is doing in the market right now.
It’s generating low-quality content.
Like I just mentioned, AI can churn out blog content with poor writing and worse research. This is taking up some of the low-hanging writer fruit…but honestly this was already being taken up by the worst tiers of sites like Fiverr, Upwork, and Craigslist.
There will always be businesses and people who want written content for less than any writer should accept. If AI fills that niche instead of somebody in India or the Philippines, that doesn’t impact serious writers with a North American or European cost of living. We should be turning those jobs down anyway.
It’s generating medium-quality art.
By comparing, pixel by pixel, different visual fields to a wide range of artworks, AI is generating reasonable facsimiles of art in various styles. Some of it is quite good, though it can be difficult to get exactly what you want with the platforms a freelance writer can afford.
Creators have mixed feelings about this. On the one hand, it can generate illustrations, ad images, and even book covers at a very low price for authors on a shoestring. On the other hand, it is carving into freelance artists’ income more than it is into writers’ — which is a moral issue for many of us.
It’s generating very, very bad books.
A few people have set AI to writing book-length manuscripts. Thus far, the results have been terrible. Although they can produce reasonably well constructed essays, the complexity of managing plot, theme, character, voice, and point of view is simply beyond their capacities at present.
That said, there are a lot of humans generating very, very bad books and putting them up on KDP. But we’re not in competition with them, so this is another thing we don’t need to worry about too much.
It’s conducting original, low-quality research.
If you ask ChatGPT or a similar program to research something, it will come up with a few facts and produce them in an easily readable format. As we mentioned earlier, that research is not great. This is because AI can only go do a super-rapid web scrape for information. It lacks knowledge and context, so it can’t test what it finds for accuracy, or even feasibility.
We can never trust AI to do research on something we don’t know about. Experts can use AI to do initial research into basic topics, and then fact-check the results — much the way a senior writer might use an intern, or a professor might give an assignment to a TA.
It’s becoming a surprisingly effective personal assistant.
From researching travel plans, to setting up meetings, to organizing finances, AI is doing well as a personal assistant for executives, micropreneurs and busy families. More on this in a bit — it’s probably the way AI is most relevant to writers’ lives this decade.
What That Means for Us
If you take nothing else from this article, walk away with the words “Don’t Panic”. AI is not the end of writing careers as we know them.
If you need more, let’s use what MBA’s call a SWOT analysis. It’s a way of analyzing the prospects of any business plan, and we can apply it here. SWOT stands for Strengths, Weaknesses, Opportunities (things that can become strengths if leveraged), and Threats (things that can become weaknesses if not avoided).
The Strengths of writers in the current AI environment come from what humans can do that AI cannot. We can approach assignments with expertise and context. We can be creative. We can handle deeply complex topics and assignments in ways AI can’t yet manage.
The Weaknesses of writers as compared to AI all lie in speed and cost. Computers are faster and cheaper than us at generating low-quality content. There’s not much we can do about that — but we can lean into how we’re the only ones who can generate content of a quality clients and readers need.
We have two Opportunities. The first is in our ability to create human connections with clients and readers. That’s something AI simply can’t do — and both freelance careers and fan bases come from our personalities. The more we do in the coming years to cement those relationships, the better we will be insulated from AI’s increasing abilities.
The second is the opportunity to use AI to our advantage, applying its power to help us write faster and better, and to reduce our own expenses as self-publishers. More on that in the next section.
The Threat posed by AI is in its exponential growth in power and capacity. As more and more processing speed becomes available, AI will produce increasingly high-quality writing. The current generative models may not materialize into a real issue within the next few years, but it’s a good idea to consider what we can do when that growth materializes or a new AI model becomes more effective.
Using AI In Our Own Writing
AI provides an opportunity for professional writers, and writers who want to become professionals. By using what it’s good at to ease our workloads, we can focus more of our time on the things human writers do that AI can’t. Below are some of the best ways we can do that:
Outlining and rough drafts
If you ask ChatGPT or a similar program to “Provide an outline for a 30-chapter epic fantasy book” or “Create a 20-chapter outline for a nonfiction book about weight loss” or “Build an outline for a 2000-word blog post about horse grooming” it will provide one. It might even be a good one. You can ask it to do so three or four times, then combine the outlines with your own expertise to create the skeleton of whatever you want to write about.
Sometimes you know what you want a paragraph to do, but you spend an hour beating your head against a blank screen because the words just aren’t coming. For writers who can rewrite more easily than they write, you can ask AI to generate the paragraph — then rewrite and edit it from there. It’s one of the better solutions for writer’s block we’ve encountered.
If you don’t like AI art, or don’t like not hiring human artists, you can still use AI to create mockups and examples from which human artists you hire can work. It’s cheap and easy to get a dozen samples, pick the ones closest to your vision, then tell your artist “like this, only with the following changes.”
If you’re out of ideas, you can ask ChatGPT to throw you lists of ideas. This works okay for your main writing, but is excellent for blog, newsletter, and social media calendars. Since many writers have trouble coming up with ideas day after day, this can be a lifesaver.
An Eye to Future
In closing, let’s look at a handful of things to keep an eye out for with AI in the coming years. We know what’s happening with AI now, but it will pay for us to be watchful for the future.
A Growing Body of Work
AI does what it does by comparing what it randomly generates to existing examples of high-quality work. With each passing day, terabytes of new examples enter the catalogues they have for comparison. Combined with increasing processing speeds, this can by itself greatly improve the quality of AI output.
The Coming Copyright War
This is a complex issue, but here’s the short version. If you tell an AI to write a story “in the style of” a specific author, it uses that author’s works in ways that might technically violate copyright. The same is true of using art. In theory, that could extend to everything AI writers and artists use as a basis for comparison.
There have already been some rumblings about this, and the first major case about it is likely going to hit the news in the next year or two. Keep an eye out for it — the decision will significantly impact AI-generated content.
Google to the Rescue
Back in 2008 through 2010 or so, content mills flooded the internet with low-quality informational articles. (I mentioned this in passing earlier). You might remember them from sites like EHow.com and LIVESTRONG. By gaming Google’s system, companies made big money off ad revenue with high quantities of bad writing. Then, in 2011, Google dropped their PANDA and PENGUIN updates. These didn’t just stop giving preference to that behavior — it specifically punished sites that used it.
It’s very likely that, sometime in the next year or so, Google (or Amazon) will produce a similar update to reduce the impact of low-quality AI-based content. Both companies have a vested interest in providing the best experience that they can to their customers (readers) and filtering out junk from being shown.
Image based on original work by Mohamed Hassan.