Every year, I spend a good portion of my time reviewing my writing skills and trying to hone them. And my favorite sources remain all-time classics like The Elements of Style, On Writing Well, and Writing Tools. I re-read them once a year, end to end. And with every reading, I feel like I get a little better at the craft.
This year, as I was going through the 2006 edition of William Zinsser’s On Writing Well, one thing stood out. Zinsser begins the introduction by describing a picture of E.B. White, hanging in his office.
“A white-haired man is sitting on a plain wooden bench at a plain wooden table—three boards nailed to four legs—in a small boathouse,” Zinsser writes. “The window is open to a view across the water. White is typing on a manual typewriter, and the only other objects are an ashtray and a nail keg. The keg, I don’t have to be told, is his wastebasket.”
Zinsser writes that White has everything he needs: “a writing implement, a piece of paper, and a receptacle for all the sentences that didn’t come out the way he wanted them to.”
White is the author of the revised edition of The Elements of Style, which he inherited from William Strunk Jr. He passed away in 1985, at the age of 86. Since that picture was taken, writing has changed a lot.
The writing implement became the computer, the tablet, and the smartphone. Word processing software replaced the paper and the wastebasket. Other technological advances have enabled us to write, rearrange, and correct text more easily. And the internet has made it easier to find source material for our writing.
“But nothing has replaced the writer,” Zinsser writes. “He or she is still stuck with the same old job of saying something that other people will want to read. That’s the point of the photograph, and it’s still the point—30 years later—of this book.”
Zinsser passed away in 2015. I wonder what he would have said if he saw the wave of generative artificial intelligence that has taken the world by storm.
AI-powered writing tools have been around for several years. But their capabilities were mediocre at best and provided minor productivity boosts to professional writers. Even the earlier generations of large language models (LLM) like GPT-2 and GPT-3 were not overwhelmingly good at writing tasks.
But with the release of ChatGPT, things changed dramatically. It has suddenly become very easy to write news reports, video scripts, articles, essays, and even books with a few easy commands to LLMs. At the same time, it is becoming harder to tell the difference between human and AI writing.
Like many others, I have been thinking a lot about the future of writing as I watch the accelerating progress of generative models.
Will artificial intelligence make human writing less relevant? Will every article we read one day be written by ChatGPT or one of its rivals or successors? Will the writer in the picture on Zinsser’s wall be replaced by a soulless piece of software trained on the collective digital knowledge of humankind?
The answer is complicated.
Functional writing will be automated
Like other technologies, LLMs will redefine the way we do things. Some writing tasks will require less human effort, and AI will probably take away some jobs. The first category of writing tasks that will be fully and mostly automated is what I call “functional writing.” The goal of these tasks is not to be creative but to convey basic information.
For example, with GPT-4 and simple prompting techniques, it only took me a few minutes to create the draft of the web copy for a software development company. All I needed was an hour of editing to get it ready for production. Other tasks such as writing product descriptions, social media posts, and presentation slide decks will also benefit from LLMs and automation.
The quality of language models in these basic tasks is enough to make writers much more productive and considerably reduce the need for human labor.
Even news reporting that provides the basic facts can also be partly automated by LLMs. Once a new event finds its way to the web, it takes a few minutes to write a news report about it. For example, I tried with both Perplexity and Google Gemini to generate a news report about a recent event, Sam Altman’s latest AI chip endeavor. Both provided me with a decent draft along with good sources (Perplexity Pro on GPT-4 was a bit better than the free version of Gemini).
With these tools, a small staff can do what previously required a newsroom with dozens of writers. News will be commoditized. Traffic will be distributed across a larger number of websites. This will drain the revenue of news websites that run ads. Some jobs will be lost.
The robots will write for the robots
Another category of tasks that will be automated are those that are meant for the software agents, or the “robots.” As search engines and social media became the main way people discovered new information, new writing jobs emerged. Here too, the main goal is not to be novel but to become visible to the algorithms that power these information discovery engines.
Take search engine optimization (SEO), for example. A considerable part of SEO involves rewriting things that already exist on the web. This is another kind of task that LLMs are very good at. They can churn out large volumes of text at a low cost. There are several examples of small teams that have managed to get search traffic with low-cost content created by LLMs. And search engine result pages (SERP) are being flooded with LLM-created text.
SEO also requires establishing long-term authority, having organic backlinks, and more. But the menial labor of writing SEO content will be mostly automated and will soon require little or no manual labor. LLMs will produce content for search engines, or otherwise put, the robots will write for the robots.
On the other hand, I’m still not sure how LLM-powered tools such as Bing Chat, Perplexity, and Google Gemini will affect SEO. I’ve already been using Perplexity for a while, and I find that it gives me reliable information from around the web without the need to click on the sources (though I still go to the source links sometimes to make sure the model is not hallucinating). If this becomes the new way people discover information on the web, it will surely reduce the volume of search traffic and the value of creating SEO content.
Returning to the roots of writing
The possible silver lining in the generative AI story is the return to the roots of writing, the kind of things that Zinsser White and Strunk taught. We have become so obsessed with writing for the algorithms, the search engines, and the social media feeds, that we have somewhat lost track of the real value of writing, which Zinsser described as “the same old job of saying something that other people will want to read.”
LLMs can’t write about the fresh human experience that never repeats itself and can’t be generalized across hundreds of billions of floating-point numbers. No matter how much data LLMs are trained on, they can’t predict or describe the joy, pain, struggles, and achievements of billions of humans in their everyday lives. That will remain the exclusive domain of humans.
Here is how Zinsser finishes the introduction to On Writing Well: “I don’t know what still newer marvels will make writing twice as easy in the next 30 years. But I do know they won’t make writing twice as good. That will still require plain old hard thinking—what E. B. White was doing in his boathouse—and the plain old tools of the English language.”
And he was right. No matter how many writing tasks the robots take away, there will always be something interesting that the robots can’t write about. We just have to look at writing from a different perspective. And good writing will still require humans who can think about their words, understand their meaning, and be able to put themselves in the place of the reader. I don’t know what wonders LLMs will make in the next year. But come January 2025, I know that I will be reading On Writing Well again because the fundamentals of good writing will remain the same.
And as we move away from writing for the robots, we will rediscover the joy of connecting with people, what I refer to as “the human experience.” We will get to directly know the people we write about and the people who read our writing. I think that in the future, there will be a market for writers who are more focused on writing for a tightly-knit community as opposed to optimizing for maximum visibility across search engines and social media networks (I’ll have more to write about this in the future).
This is the beginning of an exciting new journey for writers across the world. It has never been a better time to start writing.
If you enjoyed this article, please consider supporting TechTalks with a paid subscription (and gain access to subscriber-only posts)