Blogging in the new AI world

It seems like everyone in my professional orbit suddenly has a blog now. Even colleagues I know who have never written a blog post or an article in their lives suddenly seem to have one. Some of it has been enjoyable, insightful, and great to see. But a lot of it could be categorized as AI slop. You know what I am talking about. Em dashes every other sentence. The presence of what I call “movie trailer cadence,” which looks something like this: A short sentence. Another short sentence. A climactic point. “Two documents. One purpose. Transformed business forever.”

These things are like fireflies: once you recognize the hallmarks of AI writing, you start to see them everywhere. I am not saying AI writing in and of itself is bad. I have delivered a lot of technical documentation, strategy decks, and detailed requirements that were written primarily by AI with me acting as the editor in chief. Some of that work has been of objectively higher quality than what I would have produced manually and was also produced in a small fraction of the time. However, I do think there are some unintended consequences here that we all need to be mindful of.

I do worry about AI and skill atrophy. There is already anecdotal evidence that self-driving cars decay their driver’s skills over time. As we continue to outsource more and more of what we do to machines, we are trading our skills for convenience. In some cases, I think that is a fair trade, as long as it is made consciously.

I have made hundreds (thousands?) of PowerPoint slides over my career, one object placement at a time. Since using Claude’s PowerPoint skill, I haven’t made a single one manually. It would be an interesting exercise to find out if I still knew how.

Don’t get me wrong, I am more than willing to say “good riddance” to my PowerPoint skills. It was always a painful process I loathed. I loved the creative aspects of coming up with the flow, the palette, and the language. But the tedious nature of placing the objects on the slides and making them look less cookie-cutter was a nightmare, usually executed late at night under intense deadline pressure.

I am not willing to make that trade with my writing skills. Writing is a foundational tool for human thought, and I am just not ready to give that one over to the machines.

I also worry about the long-term consequences of the internet being flooded with AI generated content. AI’s biggest weakness from my point of view is that it cannot really distinguish fact from fiction; it is only as smart as the context it is given. If over time, AI generated content becomes the main source of context, then the output quality will continue to degrade over time. A lot has been written about model degradation when they are fed solely AI-generated inputs, and over time those models descend into nonsense. Without human generated content grounding it, AI gets more and more repetitive, derivative, and nonsensical.

A lesser but still important unintended consequence is the complete erosion of style. The more that our writing is produced by machines, the more it will all sound the same. All of this adds up to a homogenization of the written word, stripping away the value that makes each person’s point of view unique. So much of humanity and what we have is owed to the written word, and I would hate to see that disappear in the same way I would hate to see paintings and other forms of art only created by machines.

So for this blog, I decided to do something a little different. Each post will contain a little disclaimer at the top, dividing the posts into 1 of 3 categories:

Human — Every word is mine, start to finish. AI might catch a typo or suggest a minor wording tweak, nothing more.

Cyborg — I used AI as a thought partner: working through an idea, testing an argument, maybe drafting a rough pass. The thinking and the final version are mine.

Robot — I came up with the topic and the structure, then asked AI to flesh it out. I reviewed it before it went up, but the writing itself isn’t mine line by line.

I think it is fair that a reader gets to know what they are reading and if it was written by a human or not. I think normalizing this practice will go a long way in preserving the art of writing, as well as respecting the reader and informing them a bit on where the words they are reading came from.