So I've been reading about and playing around with ChatGPT and saw a piece on how people are underestimating its reasoning capabilities. So I wondered if I could get ChatGPT to play BAD (no mini-games, obviously) by describing the scenes, telling it the story and asking it to make the story choices as they come up.
I went into this thinking that that its positivity bias would probably mean it would only ever choose positive actions, in fact I thought it would fail at the very start because I wanted to give it the background of the story so just pasted the intro from the first post which contains the word sex.
But I was pleasantly surprised when ChatGPT was happy to continue.
The next hurdle was how it would respond to the language and setting the scenes. The first scene of the MC fantasizing about Josy in her lingerie, it was fine with. I really thought I was going to have an issue when Steve referred to MC has Fuckface, but once again, ChatGPT had no issue with that.
It decided to name the MC Alex when prompted.
I explained the Dik scoring system to it, and it said it understood this, I guess this remains to be seen and not sure how I test that, but I digress.
After Steve pestered Josy I asked ChatGPT what it thought of Steve, and it had this to say - 'Based on the limited information I have so far, Steve appears to be a disrespectful and harassing individual who does not respect women's boundaries.' which I thought was a fair assessment (I explain each scene whenever it changes including Steve slapping Josy's backside and the slap in the face he deservedly received.)
So far, all the choices ChatGPT made were CHICK (I'm guessing, as I don't know what it says in the code), not thinking bad thoughts, not reacting to Steve etc which is what I expected to be honest.
The first real test I felt would be the Employee of the month graffiti option, and you know I would have put money on ChatGPT opting not to draw anything, but to my surprise it went with draw a funny face.
I questioned why it chose that option, and it said - 'I chose Option 2 - Draw a funny face, because it's a lighthearted and harmless way to have fun and bond with Josy, without engaging in inappropriate or disrespectful behavior. Drawing a funny face on the picture is a playful way to express their disdain for Steve, without doing anything malicious or offensive.'
I told it that it was a good answer. It comes across as genuinely curious about the outcome, as it then said - 'Thank you. Please let me know what happens next.'
Yes, I know what you're all thinking, this guy's nuts, doesn't he have anything better to do? Yeah, probably but this is fascinating, I'm curious now whether ChatGPT will actually choose a LI and who?
I'll keep you updated periodically, though typing all the conversations into ChatGPT takes time, it also might get upset at some point or fall over, but we'll see.
Oh, here's a screenshot of the last bit of the conversation to give you an idea how I'm doing it.
View attachment 2512382
