Using Generative AI for Writing
What I have learned using AI for business and for creative projects
Ever since Chat GPT, Copilot, and Gemini became household names, many writers have wondered if AI would pose a risk to their job, art, or chances of getting published. The type of software known as generative AI has the potential to mimic and create nearly any kind of writing—though it’s not perfect yet. Through the recent AI emergence, I have worked as a writer for two different companies, and in this article I will share some of my experiences and what I have learned about using AI as a writing tool in corporate writing environments and when writing for pleasure (including this recent post.)
Relax, it’s already everywhere
Forms of AI have been in video games and on countertops for years. Home assistants like Alexa and Google Nest are early types of AI, and so is grammar- and spell-checking software. These types of AI have limited ability to think, and much of their programming is devoted to trying to figure out what you are trying to say. Without a software update, they will never get better at this or be able to provide new solutions. They don’t learn, and they don’t create.
What’s the problem?
Unlike the types of AI mentioned above, Generative AI can collect information to create new things. For instance, it could create a picture of you based on items in your grocery list or write a story about your life based on your current outfit.
That might not be a problem if you are the only person using the AI and you control what information it learns. However, instead of just learning from one person, most generative AI platforms combine what they learn from all users, which means a response you get may be based on another person’s personal information, ideas, or work. It also learns from anything it can find on the web, which may include any artwork, poetry, or prose visible to search engines. Artists, unions, and lawyers are continuously working to protect copyrighted work.
Why one of my jobs banned AI
My previous job took a hard stance against using AI to help with writing. They were concerned about who owned the material’s copyright. They were also worried that AI platforms might use sensitive, proprietary information in its learning process. And if that happened, might confidential information end up in a response to another user? There were too many questions, too much intellectual property at risk, and (most importantly) not enough of a financial gain.
My current job embraces AI
My current job sees the potential for AI. They encourage us to use it as a “writing coach” and consider it to be a tool like spell checkers and grammar checkers. We have developed some guidelines for how to safely use AI for writing and get the best results:
Ask AI for suggestions for things like alternate wording, logical next steps, examples, etc., but don’t ask it to write blocks of text for you.
Never put proprietary, personal, copyrighted, or sensitive information into the AI.
Just like with Wikipedia, fact-check the information you get from an AI search result.
My experience using AI at work
Using AI at work (Copilot in my case) has made research for my writing faster than traditional web searches. The sources are listed alongside the summary so I can go deeper for more information or to verify the accuracy of the results. The sources are typically reliable. Ultimately, my company has decided that the quality of the information and the sources our version of Copilot uses are up to our standards. So, I was given permission to use it as a primary source for a recent project. It really sped things up.
My experience using AI to write the blog
For the blog article from earlier this month, I gave Copilot more freedom. I wrote a few sentences and then told the program to elaborate on them. Ultimately, I spent as much time reading and editing the Copilot copy as I would have writing the article by myself. While I usually enjoy adding pictures after writing the article, the images AI gave me all looked like stock photos for depression medication advertisements. And when it was all done, I felt disconnected from the project and just wanted it to end. I did not enjoy writing it.
Conclusion
Generative AI is a great writing tool with emerging capabilities and improving safeguards to keep ideas and intellectual property safe. Have you used it to for writing or research? If so, what was your experience like? Let us know in the comments!
I have been concerned about AI since the early 90s when a particularly brilliant philosophy teacher raised the red flag about intellectual integrity in writing, dangers of pirating, etc. I am one person, so what I think won't matter in the larger scale. On principle, I don't use AI to assist me in writing. That way, I can say without hesitation to my publisher that this is MY work.
I think these parts are key: "Ultimately, I spent as much time reading and editing the Copilot copy as I would have writing the article by myself" and also "I did not enjoy writing it." If it doesn't save time and it's not fun, then where's the advantage? Also, many writers have reported on the inaccuracy of ChatGPT's "research." It invents quotations that people never said, and cites real sources that if you didn't take time to look at the primary source you'd never know that ChatGPT's "fact" was made up. Here's just one example from a lawyer, there are many more: https://youtube.com/shorts/60WVb1HLy8Y?si=P8SMg3KIm2IJixBe
I work at a corporation and we use Gen AI tools occasionally. Invariably, they spit out the most generic, glossy, soulless copy you've ever read in your life. Sadly, for many corporate applications, that style might be the goal. But I'd never use it for any writing that I wanted to sound like my own creative work with my own voice and personality.