Revising with AI: Infinite Possibilities
Before taking TECM 5190: Style in Technical Writing, I had never used artificial intelligence to write or revise. In the course, I spent a few weeks experimenting with HyperWrite AI, starting with an assignment to revise content with varying tones of voice. Once I became more comfortable with the tool, I also tested it to see if it could help me do research and even create copy.
I quickly learned that AI can create a seemingly limitless number of possible variations. I started with a sentence and asked the tool to revise in a certain way, such as “please make this more casual” or “please make this more friendly.” HyperWrite gave me 3 options. And if I tried again, I’d get 3 different options. If I used the HyperChat feature, I got still other options. I could also revise at the word or phrase level and learned I could easily spend 30 minutes or more considering different options for a single sentence. In that time, I could have easily just revised the sentence myself.
In The Creativity of Human Language, Richard Lederer says that “practically every sentence that you speak and write during your lifetime has never been spoken or written before in human history.” He says that humans are language inventors. Part of what makes us human is our ability to speak and write and hear words. He mentions studies showing that “it would take ten trillion years…to utter all the possible English sentences that use exactly 20 words” and a Richard Ohmann study where 19.8 billion grammatical sentences in English could be created from 25 statements which all described the same cartoon drawing of a bear and a man. No wonder I spent 30 minutes considering possible options for one sentence.
When writing, revising, and speaking, we don’t have time to consider every possible combination of words. We can only do the best we have with the time available. But AI gives us options to choose from, and it generates them more quickly than we could on our own. This could be a great tool for me, despite its limitations.
AI’s Most Serious Limitation—Fake News!
Artificial Intelligence is great at sounding very professional while telling lies. Just for fun, I asked HyperChat to tell me about the company SmartMoving, and this was the response: “SmartMoving is a moving and storage company that offers a range of services to help people move their belongings from one location to another. The company provides local and long-distance moving services, packing and unpacking services, storage solutions, and more. SmartMoving also offers a variety of resources and tools on its website to help customers plan and prepare for their move, including a moving checklist, packing tips, and a moving cost calculator. The company is based in Florida and serves customers throughout the United States.”
Sounds great, but none of that is true. SmartMoving is not a moving company. They don’t help people move or provide information on moving. And they are not based in Florida. HyperChat clearly didn’t spend much time on the SmartMoving home page and didn’t talk to anyone at the company. But it created content that sounds believable.
On a micro level, even a one-word revision may not be accurate. In a recent course assignment, HyperWrite suggested I change “businesses” to “enterprises” to make the tone more formal. That’s all well and good, but all businesses are not enterprises, so it’s not an accurate substitution.
The frightening result of AI’s tendency to state untruths with great confidence is that readers aren’t always discerning. We read something published in a book or online and assume it must be true. How can we be sure that the news we read is accurate and not fabricated by someone (or some tool) with ulterior motives or even just someone who is ignorant?
Could AI Replace Me at Work?
I used the HyperWrite – AI Writing Companion Chrome extension in my work Gmail to see if it could draft reasonable email responses. The problem: the response it created for a client promised services and timing I wasn’t willing to provide. When I draft email on my own, I get typeaheads, but these are always off, at least so far. Maybe with time the platform will improve and it will be a better option. And apparently, with time, the AI will learn to predict my content and mimic my voice. But for now, I’m writing my own emails.
I also used HyperWrite to try to generate content for a client. My task was to create a cover letter for my client, so I copied the job description and the prompt “Write a cover letter for this role” into the tool. HyperWrite gave me 3 letters, all with fake information including a fabricated degree and skills. The most outrageous line was this: “Moreover, my language proficiency in Arabic, French, and other critical languages is always desired,” which HyperWrite clearly borrowed from the bullet in the job description “Language proficiency in Arabic, French, and other critical languages is always desired.” I think the outcome would have been better if I’d given HyperWrite a list of actual qualifications from my client. But I don’t think the tool can replicate my ability to write an accurate letter speaking to the desired skills in the job description. I may someday figure out how to use the tool to shorten my writing process, but I need to make sure the output is an honest representation of my clients.
AI could be a great tool for me as a writer. It can help me research, generate content, and revise. But though it can create infinite combinations of text in minimal time, it can’t replicate my ability to choose the appropriate tone or my ability to verify accuracy. I am responsible for what I write, whether or not I use AI. It’s on me to edit and fact-check. And I believe for now clients will continue to come to me for help writing resumes.
AI can also be a great tool for me as a student. I already use the AI tool Grammarly to proofread. Grammarly helps me catch things (spelling, grammar, usage) I’ve missed, but sometimes Grammarly is wrong. It’s only a tool. Ultimately I’m responsible to produce clean, accurate content. AI may help me research for class projects if I can learn to use prompts successfully. And of course, I need to be wary of false results. AI may not be the best way to generate content as a student (unless it’s explicitly allowed) if I want to maintain academic integrity.
Benefits of TECM 5190
In Style in Technical Writing, I learned what specific linguistic variations are used to create different voice and tone. Before taking this course, I had experience writing with varying tones of voice, but I had never considered which linguistic factors made those tones different.
This knowledge (knowing specifically how to create different tones with specific linguistic variations) will benefit me both in creating my own content and in evaluating content created by AI.
The course lectures and the accompanying videos and readings gave me a valuable background in style, voice, and tone. Probably the most valuable topic of the course for me was Plain Language. Learning the characteristics of plain language and writing and revising in plain language gave me great practice in a skill I will need as a technical writer.
Advice for Future Students
Here are some of the things future students should consider to succeed in TECM 5190: Style in Technical Writing:
- Complete the readings and videos early in the week. Watch/read them once without taking notes and a second time taking notes.
- Download (and even print) all provided handouts. You will refer to them again and again.
- Visit office hours if there’s anything you are unsure of from the readings or about the assignments.
- Complete the assignments with enough time to revise before the deadline.
- If the terms phonological, orthographic, morphological, syntactic, and semantic mean nothing to you, get help early in the semester.
- Don’t overthink it. Usually, your first response is best.
- Take a look at the extra links provided at the end of each module. Dr. Kim shares lots of valuable info here that there isn’t time for in the scope of the class.