Belghast, who runs Blaugust every year, wrote today about his complicated feelings regarding some participating blogs that were using LLMs – large language models, what most people these days call “AI” – to generate their blog posts.

Condemnation was swift. Bel’s blog, Mastodon and BlueSky were filled with bloggers carrying torches and pitchforks, yelling for the villain to be dragged into the street and publicly shamed.

I also asked for the blog, mostly because I wanted to see if it was any good. My experiments with AI writing anything have been indifferent, at best. But, Bel stayed silent, and I suppose I could find out for myself by going through all the participating blogs and looking for the tell-tale signs of AI-generated text. Perfect spelling, perfect grammar, perfectly organized thoughts and flow. Better than I could ever write.

Have I used AI in Blaugust?

You betcha. Two years ago, every single day was AI generated, AI illustrated, and I had the AI generate the title and excerpts, too. Whoever is using AI now, I used it more. I was open about it. The whole point was to see how AI would handle a specific concept – games designed around the date. It was fun, I don’t regret it. I haven’t done it since, in Blaugust, but that doesn’t mean I wouldn’t do it in the future, if I could think of an angle that might work.

Back then, AI was booming; now, there’s a backlash.

The case for AI

I was thinking through what the purpose of writing is. Recording and distributing information, sure. If I want to know a fact or how to do anything, I really don’t care if a human pressed the keys or an LLM did, as long as the information is accurate (which is, admittedly, a challenge for AI). It’s at least guaranteed to be readable.

The AI almost certainly would have gotten these facts from something a human wrote at some point, and that human may have been wrong, or sarcastic, or intentionally messing with the reader; the AI can’t know that. That gets back to the accuracy, but also to what some feel is a form of plagiarism. If I write a game guide, and that is built into an LLM, and now if anyone searches for information the LLM gets it and my game guide is never seen again, I guess I’ll feel pretty bad. Especially if I make corrections while the LLM gives the old, incorrect, information.

Still, for purely informational work, AI is fine.

Busy work

There is rote writing. The classic case is writing newspaper articles about sports scores. People don’t like looking at tables of names and numbers, unless they play EVE Online, anyway. But the facts of a game, plus a few interesting details about it, can be fed into an AI, and it will write a nice article. The reporter can spend their time on more interesting work, and readers will get information more easily read.

Blogging

I’m just going to skip over creative writing, because I feel AI doesn’t really have a place in writing books for sale. I will mention that there have been for years text generators that will produce works in the style of any given author, that long predated ChatGPT. I haven’t tried any of them, because I don’t care to read what they output. The works would be, at best, a parody.

Call me Ishmael. Or don’t—call me anything you damn well please. I was drifting through the smog-choked streets of Manhattan, my brain buzzing like a hive of angry bees. It was one of those days when the city seemed like a sprawling madhouse, and I felt a desperate urge to escape the concrete jungle and its legions of lunatics. Moby Dick (in the style of Hunter S. Thompson)

Okay, I take it back. Having ChatGPT write out the first chapter of Moby Dick in the style of Hunter S. Thompson was pretty fun, actually.

Anyway.

I don’t blog to inform anyone of anything. If anyone wants to find out anything, this blog is the last place they’d go. I don’t play games well, and when I do, I mostly mess up. The games I do play have been well-documented everywhere else. I can’t even get my family or friends to read the blog, and my most popular posts have been about cheating at Wordle and stuff like that. So: not informing anyone. Don’t need AI for that.

Rote writing – busy work? Nah, there’s no mandate for me to write anything. Nobody is looking for this blog for scores or anything. My most read work was when I was writing about the adventures of my old EverQuest guild, Crimson Eternity, and that’s because people were interested in what our guild was doing. But, that’s long in the past.

No. I write because I am sending messages to my blog’s number one reader – me. Future me, actually. I read the old posts and remember who I was when I wrote that. Heck, most of the times I don’t remember the event I was writing about, or it reminds me of some game I played that I really liked, or that time I worked for Massively. So most of the time, when I blog, I’m blogging for me.

I don’t need AI for that.

This is my blog. If I want to use AI to send future-me a message, I will. If some other blogger wants to send messages to their future self using AI, more power to them. If they say, pay me money because I totally wrote this perfect piece of prose using my own two brain cells, cool. If someone wants to pay them for that, yay, they both win.

AI-using blogger, you won’t read this. But just know, I’m rooting for you. This year you use AI, maybe next year you’re confident enough to use your own words, maybe the year after that, building on this foundation, you’re earning money with your own thoughts and ideas.

Or maybe you don’t. It is your life. I’m not going to tell you how to live it. As long as you don’t hurt anyone, do what you want.