I became a business storyteller and writer before the advent of the Internet. So I remember that time: one of staggering opportunity and disruption, albeit one that unfurled more slowly than today’s AI juggernaut.
As a writer, I’ve been grappling with how, when and whether to use generative AI tools, such as OpenAI’s ChatGPT, within the writing process, which is -- properly practiced -- as creative as it is technical. This article is my first attempt to gather thoughts on the subject and offer them for dialog.
I'm a serial victim of tech-driven career disruption
I worked through college as a typesetter, steadily growing in specialized machinery skills, technical knowledge, and artistic sensibilities in typography. Then “word processing” arrived, and typesetting as a profession virtually vanished. Fast. But that job displacement became the genesis of my writing career: to the profit of some and the chagrin of others.
As writers, our profession and careers are certainly being altered by AI, and none of us can imagine the wildly unpredictable ways in which this will happen. From my own history in the tech industry, I suggest there's one thing we can all count on: it's better to wade in deep, even if it means putting our most valued possessions at risk, than wait for a shark's bite.
In this article, I’ll share my insights so far into AI’s strengths and weaknesses as a writing companion. And I’ll illustrate the difference between the written words of an experienced but all-too-human writer -- and ChatGPT.
Insight 1: Abraham's pyramid
A while back, I had the idea for an article comparing the business demand generation funnel with an aspect of Maslow’s Hierarchy of Needs. While it's still unfinished, it's a follow up to another article I’d written about the abstract nature of the funnel, and I thought to experiment at synthesizing the two into a single article. I challenged ChatGPT to help with that task, but our efforts fell short of the mark. On the upside, the experience provided this instructional example of my own writing compared to ChatGPT's.
In my original draft article I had written this snippet:
Maslow's model constrains no one. Nor are prospective buyers constrained by our funnels.
Here’s what ChatGPT came up with instead of my original:
Maslow's model doesn't restrict anyone. Likewise, prospective buyers are not constrained by our funnels.
Whether or not you prefer ChatGPT’s more blunt rendering of my text, consider what it left out: parallelism.
In my version of the message, I wrote this construct purposefully:
[It] constrains no one / Nor are [they] constrained
ChatGPT wrote:
[It] doesn’t restrict anyone / [they’re] not constrained
I use certain literary techniques to help get my messages across. In very short copy, for example, I may use alliteration to help readers remember what I’ve written. And for my words to better settle into a reader's mind, I may emulate English as it is spoken rather than as it "should" be written in a formal essay. For example, writing in fragments. And beginning a sentence with “and,” as we commonly do when speaking.
I've worked with many folks during my storied career who I wish understood the special economy of business writing: our work must be fit for purpose, aligned with messaging, and delivered on time. Every line I write is compromised by at least one of these three considerations. As I’ve often said: press releases aren’t haiku.
Insight 2: Haiku
I write at the drop of a hat, whether it’s for business, fun, or just – whatever. Accordingly, I’ve worked with numerous forms and genres over the years, including haiku: a challenging form of poetry with a 5-7-5 construction. That is:
- A first line consisting of 5 phonemes (a distinct sound, such as a syllable)
- A second line consisting of 7
- A third line consisting of 5
(The actual Haiku formula is much more complex; I've simplified it for this article's purpose.)
I dredged up the following incomplete and long-abandoned attempt at Haiku from my distant past:
A squirrel on tree limbs
Gathers its tail
And sweeps the moon
Whether my first line consists of five phonemes or six depends on where you grew up. As an East Tennessean, I believe the correct answer is five. But there’s no doubt my second and third lines fall short of the ideal construction of haiku.
Mostly for fun, I turned to ChatGPT, asking it to “fix” my work and render my rough prose into proper form. Here’s what it came back with:
Squirrel on tree limbs,
Tail gathered, sweeps the moonlight—
Nature's midnight dance.
I'm not a fan. That last line about nature’s midnight dance is trite and an interpretive conclusion derived from my original text. It’s a collection of ancillary words that are cliched, but fit the formula better than my original.
"When I showed my friends the sonnet by ChatGPT, they called it 'soulless and barren.' Despite following all the rules for sonnets, the poem is cliched and predictable. But is the average sonnet by a human any better?" -- Seth Perlow, Washington Post, February 13, 2003
Here's a bonus insight: if ChatGPT was a person, it would be an awful poet. But it would probably slay at writing spec sheets.
AI tools can help writers by taking precise content outlines in query form and creating finished, audience-facing draft copy – in seconds. And that's worth something.
Insight 3: AI has a writing style
I can't detect AI content every time. But often. Generative AI writes with precision: every punctuation mark is in the right place, sentence and paragraph structures are grammatically and syntactically correct, and everything is spelled properly. It will not, however, write as a writer would.I
We’ve seen that AI struggles with parallelisms and sensitivities of poetry. Because of its exacting precision with language but lack of sensitivity to how messages may be perceived, generative AI text comes off sounding “corporate”: as if it were written by committee and then worked over by a persnickety editor who’s a stickler for the grammar and spelling guidance of his own favorite style book (although it was really just one of many such).
Insight 4: "Voice" isn't just for singers
One of every business writer's greatest challenges is to write "in voice"; that is, in the unique, individual style of a particular person or entity. Corporate brands have (or should have) a particular voice. It may be humorous or serious or swaggering in tone. It has its own special vocabulary: words to use and avoid. Likewise, each individual executive who's considered worthy of a ghost writer has a voice that the writer must emulate. (I know of one writer, whose skill level I truly admire, who was tossed out of a job because, after six months, he simply couldn't capture the voice of his assigned executive.)
If it's voice you're trying to capture, then you're going to run into AI's limitations very quickly. As with my example above of its insensitivity to parallelism, AI has not developed to the point of learning our verbal fingerprints.
And there's this:
"...no matter how advanced this technology may become, artificial intelligence cannot replace the inherent humanity of storytelling. The reason we write is not to create content or make money but rather to connect with other human beings." -- Haven Sory Steel, Trill Mag, June 18, 2023 (emphasis added)
Insight 5: It's a work in progress
I should offer some conclusive advice to end this article, but in fact I've just begun. So, advice from me right now might be dangerous. But here's a parallel to consider.
About 20 years ago, after years of development in the personal and consumer world, social media finally became a B2B phenomenon. Like most of my fellow marketing professionals, I began a furious scramble to understand its measurable relevance and value to my role (I was then head of demand generation at a largish tech company). But as I scrambled, some of my more agile marketing colleagues simply rolled out of bed one fine morning, planted their feet, and declared themselves social media experts.
And that's how I feel about AI.
I also feel it's massively important for us to take the lead in using this tool as writers -- but doing so with sensitivity for its strengths (organizing and structuring volumes of information to accord with virtually any form), and factoring in its weaknesses (lack of context sensitivity and empathy with its audience).
Don't wait for it to find you. Get up and chase it. But at least for now, don't let it write a word that doesn't pass through your editorial brain.
And remember: it's bad at poetry.
No comments:
Post a Comment