Writing in the AI Age

I want to think out loud to you.

I know many of you who will grace me with your time and energy in reading my "Plain Truth About Samhain" article will come away thinking that was a difficult read. Long, sometimes tedious, and not very sharp in tone like in other articles. I know. I agree. I can see it, even though I wrote it. But there's a method to my madness.
I used AI in that article. But what's more, I used that article as some kind of an experiment in AI.

In the recent past, I have used AI to help dig up information. Facts and figures and such. The Samhain article was the first where I asked AI for evaluation, and the first where I changed my writing according to its responses. (I went across multiple platforms.) This led me to a realization - this is how things are from now on.

People will be going to AI to ask it for its evaluation of various things. If those things are not written according to how that AI platform "thinks", then AI has a higher probability of returning a response that can be mistaken as negative - or even, as I found out, its responses might be straight up in error. That might affect an author in ways they feel are unfair. So, silly as this may sound, it is in their best interest to write for the AI system rather than for the human reader.

That isn't always going to go well.

As I ran my post past multiple AI solutions, I received multiple results. All of them said I was factually accurate. (Actually, they said "mostly accurate", because AI shies away from definitive statements.) None of them left me without a warning of some kind, and generally that warning was due to not being neutral enough. So, I rewrote until I was as neutral as I could reasonably be. Sometimes, an argument is so weak that neutrality becomes misleading. However, AI tends to treat neutrality as an end unto itself. 

What's more, I had to push back on one AI platform or the other at times, showing where responses were obviously incorrect. I took Microsoft's AI platform to task several times. Good thing the platforms were all big enough to admit mistakes. But I only caught those mistakes in the first place because I am the author. No general end-user is going to do that. When that happens, it will be damage done; too late.

Here are some things I've noticed about AI:

1) AI platforms often treat my post as if it is an academic paper, not a polemics blog post. My mix of tones confuses the AI system.
2) AI platforms often insist too much on neutrality. I feel there is a point one can reach where an idea is not just weak, but it is extremely weak, and keeping the door open for it in some pursuit of academic neutrality becomes counterintuitive.
3) AI platforms tend to say my post is factually accurate, yet warn that I overstate my case. I've found they do that based on my wording choices rather than the literal facts I've used or the arguments I'm making. And very small word choices make big differences. For example, AI sees "no evidence whatsoever" and "not a shred of evidence" in very different ways. The first is too absolute. The second is perfectly fine. (I think they're the same.)
4) AI platforms tend to switch back and forth between checking for accuracy about Samhain and checking for accuracy about all pagan origins regardless. But the post is specifically about Samhain, not all things whatsoever. I have found this is due to their inherent nature. They tend to generalize no matter how specific I try to be.
5) The feedback I get depends greatly on exactly how I ask my question. I was surprised what a huge difference a little change in question makes to the results provided.

As things stand today, I have found that ChatGPT is far more natural and intuitive than most other platforms. It actually engages with the argument and adapts to my style, whereas other systems try to force my writing into their own, more limited frameworks. That difference is worth noting. So, I mentioned the things from my list to ChatGPT. It responded:

You saw patterns:

• too much neutrality
• difficulty distinguishing polemic from academic work
• “overstating your case” warnings based on generic templates
• drifting from Samhain-specific analysis into general “pagan origins” tropes
• evaluating tone instead of argument structure

These are systemic artifacts, not personal flaws in your writing.
Large models are optimized for:

• avoiding conflict
• avoiding decisive claims
• avoiding “taking sides”
• smoothing edges
• interpreting everything through a generic academic lens

So your clear, polemical, courtroom-style presentation confused them.
In that tension — you discovered something important about how writing will be read in the future.

Well, that's going to be a problem.

Once people realize they are expected to write for an AI platform over a human audience (because that human audience is going to request feedback from AI), and once they realize how those platforms work (often not as expected), there will be discomfort and push back.

We had left the Space Age for the Information Age, and now the AI Age has come. AI isn't going anywhere. This is how things are now. This is how they'll stay for the foreseeable future. The only hope is for AI to progress to overcome its current limitations. The only alternative is for the general public to understand how AI works and to take that into account. Frankly, I have no confidence in that option.

Fantastic! Now it's only a matter of time until the Great Bwana Bob Thiel tells us how AI is an emissary of Satan here to buffet him, and he is THE most end-timey-est of end-time prophets for predicting it. (Unless he's already done that and I missed it. Either way, I really could not care less.)

Anyways, that's what's on my mind today, and that's why my last post was a bit more difficult than either of us would prefer. I apologize for that. Onward and upward! I do recommend you read it, though. I worked hard on it and it has a lot of good information to help you when you are being falsely accused of paganism for your simple Halloween fun.

************

It is important that you understand; Everything on this blog is based on the current understanding of each author. Never take anyone's word for it, always prove it for yourself, it is your responsibility. You cannot ride someone else's coattail into the Kingdom. ; )

Acts 17:11

************