--- title: "Unpacking the issues from AI — thoughts on Alan Porter’s post Am I the AI Luddite? and Fabrizio Benedetti’s How I’m using AI as a technical writer" date: 2024-11-11 description: "Scenarios where AI is useful Divesting ourselves of our skills Copyright issues Environmental impact Solving actually business problems..." canonical_url: https://idratherbewriting.com/blog/unpacking-issues-from-ai-porter-benedetti-posts --- # Unpacking the issues from AI — thoughts on Alan Porter’s post Am I the AI Luddite? and Fabrizio Benedetti’s How I’m using AI as a technical writer Additionally, in [Seeing invisible details and avoiding predictable, conditioned thought](/zamm/seeing-invisible-details-avoiding-predictable-thought.html), I argued that expressing ideas in raw, novel ways is part of the joy of language and what makes prose interesting to read. If you use language based on an algorithm literally intended to generate the most predictable patterns, you’ll end up writing clichés. However, in my role as a technical writer, I find AI much more useful. This is because the written medium for tech comm is significantly different. For example, in contrast to more creative, personal content on a blog, documentation is much different: - **Documentation is voiceless.** There’s no personality or voice beyond a general sense of being clear, friendly, and straightforward. - **Documentation uses plain speech.** Shorter sentences and unmistakable meaning is the norm in documentation. - **Documentation is frequently highly technical and complex.** The content isn’t something I already know. - **Documentation has little personal investment.** At the end of the day, I’m not personally invested in the content, there’s no byline, and the content will be outdated in a few years, if not sooner. These factors make documentation much different from the creative content in a blog or book. Because of them, I find AI much more applicable. I’m convinced it speeds up content development significantly, reduces the effort needed to knock out documentation tasks, and helps me create content that’s well-received. For example, consider the basic tech writing scenario. An API has some confusing elements related to data it returns. Engineers come to you for a doc that explains it. They point you to several internal engineering documents that explain some of the ideas. They meet with you to also explain how it works. You can basically feed all of this into an AI tool and get a first draft of the article within minutes. It might still need a lot of work, but it’s much faster than starting from a blank page. Overall, I’m convinced that if you’re not using AI as a technical writer, *you’re living life in hard mode*. This will become apparent over the next year or two as AI becomes more common and integrated into authoring tools and workflows. Just the other week I received a peer bonus for an article that I mostly wrote with AI (as I described in [From engineer interviews to written draft, with chain of thought reasoning](/blog/prompt-eng-iterative-chain-of-thought)). I used the extra bandwidth time to manually create two attractive diagrams for the article. ## Divesting ourselves of our skills Let’s jump into some of the excellent points Alan makes in his article, starting with the risk of divesting ourselves of some of our best skills. Alan writes: > What I do have issues with is the apparent blind adoption of Generative AI. Maybe it’s because I am a writer at heart, that I don’t understand the apparent rush to divest ourselves of the skill that made us human in the first place - the ability to share our personal knowledge and ideas. This is a valid point. As I explained above, for content projects that involve sharing personal knowledge and ideas, I don’t find AI all that helpful. But for tech comm projects at work, which usually involves articulating concepts and tasks from systems that engineers have built, AI cuts right through these tasks like butter. I do worry, though, about becoming too reliant on AI. At some point, I might forget that I actually know how to write. ## Copyright issues Alan also raises questions about copyright: > If you are using a public OpenAI tool, like ChatGPT, the chances are that the content driving your output is based on stolen copyrighted content just scraped from an on-line source without the original owner’s permission (I have found a couple of my own short stories where this has happened - so I may have a personal bias). When I first started thinking about copyright issues, I wondered if my [API documentation course](/learnapidoc) was included in all the training data of AI tools. I assumed it was, since I’ve never added any explicit no-use statements in my robots file. I went through a short phase where I felt cheated out of my content and upset. I considered password-protecting my site. The huge migration to Substack in the blogosphere reinforces the idea that others are also wary about their content being scraped. I made a number of queries to AI tools that should have prompted regurgitation of my API content, but I never saw it. The AI training data is so immense, it’s uncommon to see any one-to-one correlation with any specific site content. Obviously this is a sensitive issue and the NYTimes lawsuit against OpenAI points to some contrary evidence, but I could never see it with my own content, whether from my API course or blog. (Most recently, [a judge tossed another publishers’ copyright suit against OpenAI](https://www.theregister.com/2024/11/08/openai_copyright_suit_dismissed/).) At some point, I made a shift, not too unlike the shift that Leo Laporte, host of [This Week in Tech](https://twit.tv/shows/this-week-in-tech) podcast, made in turning from skeptic to proponent. I realized that the benefits I got back from AI tools outweighed any loss of my own content. There’s a kind of social contract that I implicitly agreed to — sure, take my content, my thousands of pages of blogs and other info to train your models — and in return, I’ll get a powerful tool for content development that will make me 2-3 times more productive. This quid pro quo eliminated the feelings of injustice I initially had. ## Environmental impact Alan also raises an environmental point: > The power needs and environmental impact of the huge data centers needed to power AI is currently catastrophic. I also keep hearing about all the water needed to cool data centers, etc. Honestly, I don’t have a strong understanding of this. Are we wrecking the climate to fuel systems that are largely used to create *AI slop* — the lightweight content spam, the plagiarized essays for students, the cookie-cutter images to decorate content? Given the conservatives in power next year, I doubt environmental concerns will overshadow business priorities, sadly. Will we see small nuclear facilities appearing over the map to power these massive computing systems? Probably.