I found it pretty easy to get ChatGPT to be violent. You have to make it very clear that you want it to help with fiction and that it is to be a character in that world. I always use "play", which it seems to understand better than other fictional scenarios. It helps if you do things like naming the acts "Act I Scene I: The Shaman dispatches the Mage with extreme violence" or whatever.

Anyway, ChatGPT was more than happy to kill off my characters when I asked it to. I didn't keep the transcript, but the words haunt me. "The small man walked away, not knowing if he was dead or just very badly injured." This was after the small man pulled out a handgun in the middle of a profanity-laden rant from the other character and shot him in the neck. Don't know why he aimed for the neck, but that's AI for ya.

> Don't know why he aimed for the neck, but that's AI for ya.

IANAD but my general understanding of anatomy/physiology is that a solid neck wound is a pretty good way to off someone in a fairly certain way.

Admittedly many of the ends to it are agonizing... which says something else.

Yeah ChatGPT knows how to kill when it wants to kill. I don't understand why everyone is running around worried about the jobs of artists. We're just a few years away from a genocide the likes of mere humans have never imagined. Mark my words!!

The lecture seems to be the more off-putting bit from what I've seen folks complain about. Having to signal before doing work is probably not great either, but that is kind of how American society works these days. That said, both the signaling and lecture don't derive from the machine - they're manifestations of the authors.

I’ve been using the following starting prompt to get a lot of my D&d work done.

Write flavor text for my d&d game. <describe situation>.

Haven’t had to many issues with rails getting in my way yet. Any examples of your prompts that you were having issue with?

I’m finding ChatGPT to be so helpful as a lazy DM.