A (pointless but funny) ChatGPT Experiment

I was playing with ChatGPT earlier. I do this quite a lot. This time I wanted to see how well it could act with different personalities in reply to a social post, but then things got out of hand.

The prompt, in summary, was giving ChatGPT a (made-up) post from social media and then asking it to write replies from five different simple personas. I didn't give it much to go with persona-wise (a bootlicker, an angry person, the OPs boss and a couple of salespeople), but I thought the generated responses were very interesting.

Particularly the AI-generated social media spat that it led to.

This was my prompt

And this was what it generated

I then asked it to add a little more fire to the "angry person"

You'll see that it took a couple of attempts. The first was weirdly formatted, as it had replaced the "uck" in "fucking" and the "i" in "shit" with *s, and because ChatGPT uses markdown for formatting (which makes everything between 2 *s bold, as WhatsApp does), it confused itself.

Then I got started to get creative...

Now, 3) Slimy Salesperson replies to 2) Angry... but needs to be nice in case Kevin sees what he says

I know I'm stoking the fire here, but let's see how angry we can get 2) Angry.... (I've now assumed they're male, for some reason)

Slimy Salesperson has dropped their guard. This may be my fault (apologies for the language)

That weird formatting issue rose it's head again.

Let's get Kevin (the OP) to close us out

Kevin has focussed on the argument but has forgotten about Salesperson1 and Salesperson2's pitches.

Oh no - we've lost Kevin. Man down!

That was the end of my experiment. I had to go and have a lie down after this.

Joking aside - the power of ChatGPT is crazy. Whether it's used for good, evil, or just the overgeneration of content (I suspect the latter, in the main - kind of mild evilness), it's hard not to be impressed with its capabilities.

Me, I'm trying to do some good stuff. But often get distracted. The above is a case in point.