Artificial Intelligence, Writing, and Editing

We live in interesting times. Artificial Intelligence (AI) technology is developing rapidly, and there is intense speculation and worry abound around its impact on work. Editing work is subject to these worries like any other job. So let’s do some speculating today!

We think that editors don’t need to fear the machines yet, but we do have reasons to fear our fear. Let’s unpack that claim.

What are the important AI systems for editors? There are two major classes of AI technologies that will surely impact the editing market. First, there are AI engines like GPT-3 that have gotten good at generating natural-language text. Second, there are AI-driven engines for editing text, which are more speculative. Today, we’ll focus on the GPT-3, otherwise this post might get unwieldy.

The future of writing and editing? Photo by Markus Spiske on Unsplash.

There’s a well-known article that was written by GPT-3 for The Guardian. It’s interesting. But the details matter. GPT-3 was given a fairly substantive opening, and then generated eight versions of the article, which was stitched together by the Guardian editorial staff. As the post-script to the article states: “editing GPT-3’s op-ed was no different to editing a human op-ed. We cut lines and paragraphs, and rearranged the order of them in some places. Overall, it took less time to edit than many human op-eds.” Editing was still absolutely central to the process.

What does this incident suggest? GPT-3 might impact the editing market positively, driving up demand for editors, because it will increase the volume of text that needs editing. It might be true that GPT-3 can write about as well as the average human, but editing work is not mainly focused around average writing—it’s about making the writing more than acceptable! If we drill down a bit, GPT-3 might decrease demand for copyediting and proofreading, since here its computing power gives it an obvious advantage over human writers, but it is likely to increase demand for developmental and stylistic editing. This is because GPT-3’s output is coherent at the sentence and paragraph level, but starts getting wonky for larger writing structures, as suggested by the making of the Guardian piece. Perhaps in the near future clients may start a writing project with AI-assisted writing and then may find that they need editorial intervention to make the text more coherent at large scales like arguments, chapters, rhetorical framing, and through-lines.

So, where editing is concerned, there may be changes coming in the near future, but worries about the collapse of demand for editors is premature. One take-away is that editors might want to focus their professional development on high-level editing.

However, there is one worry around AI and editing that we take very seriously. We think that anxieties like “machines will take our jobs” might trigger editors to prematurely make themselves cheaper, triggering a race to the bottom. The antidote to this possibility is to look under the hood of these AI systems and have a working knowledge of what they can and can’t do. We think that both dismissing AI and over-inflating its powers are errors. There are real worries, but we need to be measured about them. As usual, more education will help everyone. Stay tuned as we explore these ideas in a bit more detail in future posts.