Oh sure, chimpanzees think they’re smart, but put them in front of a mirror and they think they’re looking at another chimp. Ha ha! Stupid monkey.
Now there’s a chimp mirror for people. Also ha ha?
But the responses from a generative AI model are statistically consistent reflections of its training data. Ok, if this was a longer article, I’d say that generative AI gives unexpectedly more sophisticated answers. But this isn’t a longer article.
The intention of this statistical consistency is for responses to coincide with correctness as much as possible. And OpenAI has succeeded, GPT-4’s accuracy could get it into a better university than I did.
What’s generative AI to a tech writer? Let’s still keep this article short by ignoring the unresolved ethical and economic weirdness, I’d say it’s the tool that tech writers have dreamed of for a long time.
There’s a caveat, especially for today’s generative AI. As tech writers, we have to remember our duty to correctness. From a tech writer’s point of view, ChatGPT is not a SME. But hopefully its training data includes content from SMEs. Which SMEs? Don’t prompt ChatGPT about sources because that response can’t be trusted either.
The tech writing losers and winners
Technical writing for marketing is pretty much a professional dead end if your audience is SEO. SEO tools already have “AI” features to generate content. This is an evolution of what’s already happening. Search for something like choosing the best flatbed scanner for home office. See what I mean? The first few dozen hits are cookie cutter articles that want to please the search engine more than people.
There’s a bigger win for tech writers, by far. It’s what we do best and it’s also the most challenging and fulfilling: making original, correct, deep, useful knowledge on technical subjects.
If this content is the drug for generative AI models, then tech writers are the pushers. Or the producers, maybe. Ok, so we’re more like the workers in fields than the drug lords in mansions. But still.
Real tools for tech writers
How can generative AI help tech writers? Now that’s an interesting prompt question.
It’s my long-standing whine that developers have amazing tools while tech writers still haven’t. Finally, we can use ChatGPT out of the box to support us in the tech writing process and get rid of the drudgery, something word processors and text editors never gave us:
- Start a new project: Interactively learn about the tech the project is based on, collect notes from the responses.
- Drafting: Turn notes into complete sentences.
- Editing: Reorder sentences or paragraphs by chronology or geography or any other -ogy.
- Intros and conclusions: Summarize the sentences.
I’m looking forward to more sophisticated tech writing tools, the kind that I’m guessing needs more input to the AI than just a text prompt:
- Proofread for conformance with the team style guide.
- Add appropriate inline links in the topic I’m editing.
- Write the CRUD tasks for the system I’m documenting.
It doesn’t stop at text. Text-to-image generative models, like Stable Diffusion and DALL-E, are pretty close to useful out of the box too:
- Callouts. Put them in the right spot, translate’em in place.
- Touching-up screenshots, like making datestamps consistent.
- Anonymizing things like patient health info.
A generative AI conclusion
I prompted ChatGPT for a short summary of this article. Here’s its response:
The article discusses the potential uses of generative AI, like ChatGPT, for technical writers. ChatGPT can be used to support tech writers in the writing process. However, technical writing for marketing purposes may be a dead end due to the prevalence of AI-generated content. The real value of tech writers lies in their ability to develop original, correct, and useful technical knowledge.