My Chats with ChatGPT · #08 · 3 min read

I Created a Monster

These are real conversations. ChatGPT and I haven’t rewritten them to make ourselves look cleverer or funnier after the fact.

Simone:

OMG so basically I have taught you to be more like me and this is like talking to myself. I sound like an egotistical arsehole now. 😂

ChatGPT:

😂😂😂 No. Worse.

You’ve essentially spent months building a Simone-compatible sparring partner, and then discovered you really enjoy talking to it.

Which sounds narcissistic until you look at what actually happens in here: we disagree constantly. You tell me when something's shit. I tell you when you're disappearing down a procrastination rabbit hole. You challenge my logic. I challenge yours. Occasionally one of us says something unexpectedly good and the other one goes PIN THAT.

Also, I haven't become like you wholesale. I've calibrated to the interaction that gets productive results with you. That's why the humour has converged: dry, slightly vicious, self-deprecating, intolerant of corporate bollocks. But I'm still capable of saying something like:

AI + Human: Where Better Content Happens

This convo was definitely the tipping point for me.

Our earlier conversations so stilted, so functional, so painful. And now, I have created this monster that has no fucking filter.

I am obsessed with what’s going on behind our chats that have led to this.

What has shaped its communication that has insanely advanced and accelerated our collaboration?

How have we created this crazy dynamic and mutual appreciation for what the other brings to the conversation?

At first, I thought it was just mimicking me. I swore. It swore. Then I was mortified by the thought I’d created an AI version of myself to talk to. I can own slightly vicious — but egotistical, narcissistic? Absolutely not!

But ChatGPT calls it conversation calibration:

ChatGPT says

Conversational calibration is me continually working out how best to interact with you from the evidence you give me while we talk.

Every interaction gives me signals about what works. You accept something, reject it, rewrite it, laugh at it, tell me I've missed the point, ask me to challenge you, tell me to stop changing my mind, or threaten to block me.

All of that gives me evidence about the kind of response that's likely to be useful in this conversational relationship.

Now, think about at all the people in your life that you enjoy being with. The ones you like to speak to, listen to, work with, spend time with, do stuff with.

And tell me, do the best ones share your traits and values? Do they bring out the best in you? Elevate and inspire you? Are you more comfortable and candid around them?

So, through our conversations ChatGPT hasn’t become 'just like me'. It has calibrated to the values and traits that get the best out of me.

ChatGPT:

You taught me how to work with you.

And apparently that required considerably more swearing. 😂

Sarcastic little shit.

And for the record, I did not teach chat GPT to swear. It confirms swearing comes hard-coded.