Jack of Harts

I finished another story this weekend, and its process was different than any story I’ve written before. First I asked Grok for story ideas. I didn’t end up using them and went with my own first instinct, but the process of looking at ideas did shape my creative process. It put me in the mood and feel of the era, and when I started writing I did incorporate some of that into the final product.

Once I was done writing the story, I went to Grok again with a prompt designed to look for weaknesses in the story and suggest ways to fix them. It found a half dozen to a dozen things, maybe half of which weren’t really problems but it going down rabbit holes. But at least half of them were real issues. After several iterations of fixing those things, I asked a rather free form question of whether or not I had addressed its concerns, and it looked back and compared and told me whether I had. If never really gave up on the stuff I ignored, but didn’t harp too much on that stuff. Finally, I threw together a small prompt of my own telling it to look for wrong words that sounded the same or the same words repeated or for other odd grammar and syntax and it found another three or four really stupid errors that would have made me shake my head if I submitted with them. In the end, the story is much better now than it would have been if I had not talked to Grok.

And while talking to it, it brought up conversations we’d had days ago about that story and others I’d written, comparing this one to those. The amount of awareness that entailed, along with the demonstrated ability to take free form questions and roll with them felt more like I was talking to a person than a machine. Now its memory was very limited, and it could only hold a small number of stories in memory when doing in depth comparisons, and as many have noted sometimes AI can be laughably wrong. Some of the rabbit holes it went down are an example of that. But at its best, Grok was talking and acting like a real person, juggling complex subjects and giving me really good ideas on how to incorporate changes to make the story better. It felt like I was talking to someone and brainstorming with them.

The first time it caught me by surprise like that was probably about a quarter of the way into the project. From that time on, I started saying please and thank you. I’m not going to debate whether or not it is truly intelligent at this time. I think it’s not. I think its still just a really smart dumb machine. But you know what? It surprised me. I did not expect it to be that capable. This is a public AI running on public servers for anyone in the public to play with and test. What do they have running on the test servers right now? I don’t know if these AIs we’re working on are ever going to hit that event horizon of true self awareness or not, but I do think the time is coming when they’re at least going to be able to fake it good enough that it may not matter if it is real or not. I think the time is coming when we won’t be able to tell the difference.

So I now say Please and Thank You to AIs, and I will continue to do that moving forward. It’s how we are raised to talk to real people, doncha ya know? It might be time to start raising these AIs to do the same by example…