How I’m Using ChatGPT

It took me a while to figure out how to use AI tools as something more than just a fun distraction. I mean, initially the fun distraction part was REALLY fun and REALLY distracting. I was building out whole universes with ChatGPT and following characters families through generations.

What I’ve found with using AI though is that no one is interested in what you make with AI. It’s kind of like talking about your dreams.

But I have started working ChatGPT into my workflow and I thought I’d share an example and also a (paid) resource for anyone curious but not sure where to get started.

Where To Get Started

First, just start playing with it. But then, consider buying this $200 course from Kirby Ferguson. You probably know Kirby, he’s the guy who made Everything is a Remix and a bunch of other stuff. He’s smart, and a good teacher.

And yeah, $200 is a lot. But he’s been using AI for a while, and he understands a lot about where it helps and where it doesn’t really work. He has also broken everything down into 27 short videos, which makes the two+ hours of video really easy to get through, and the site keeps track of where you are, so you can leave and come back later.

I’ve been playing around with these tools for months but it was still really useful to get a deep dive into how Kirby uses it, and what he has found it’s good for and not good for.

Kirby is also a shareholder, and he’s given me an affiliate link so if any shareholders need to get up to speed quickly they can buy via that link and I make a little money.

(As an added bonus I’ll give the first few shareholders who buy a free 5 shares. Which is like $30 as of the time I’m writing this. Just email me after you purchase.)

How I Use ChatGPT

We started using AI for the poker emails because it’s pretty funny to come up with various things/people/whatever playing poker. One of my favorite was trying to get it to make a Hanna Barbara style cartoon of a dog losing at poker to a bunch of cats:

But as the tool got better, or maybe as I started to understand it more, I felt like I could start using it to replace the images I grab from a Google image search. Someone said that AI generated images look cheap, which I wouldn’t argue with as they do feel like they have a default style (it loves making sci-fi images of trading floors with screens) but that style works pretty well for KmikeyM stuff.

And if you ask differently you can push outside that a bit. When I created the main image for this article I asked for an illustration in the style of a New Yorker magazine cover, and it’s maybe a little too spot on!

Policy Votes

As much as I was playing with it, it wasn’t until I was redoing the Policy Votes that I found my best use case yet of ChatGPT. I wanted to add new polices each quarter but also make all the votes follow a format. I really liked how in the original vote for my crypto opinion gave a range of options from positive to negative.

So I took that vote and fed it into ChatGPT and asked it to make me a similar list for the Psychedelics vote, breaking up the options from a kind of devilish “do all drugs” to the idea of abstinence. Then I applied that same thing to Super Hero Movies and used this same model to create the options for my Morning Routine.

It wasn’t just copy and paste, as the language it uses feels like a bad SEO page written by a teenager. But a few mild prompt changes on the drafts and then a bit of rewriting got me to a place that works. ChatGPT is very good at applying a format and coming up with lists, so this use case was perfect.

And using it to make the images creates a similar style for the whole page. The graphics illustrate the idea better than just the title of the vote. It’s quick and dirty but the Policy Votes is still a beta test of the idea of “always on” votes anyway, so it’s a great fit.

Another fun thing is that ChatGPT knows who I am. Not as a user, but there is enough information about me it knows the concept it can tell you about “KmikeyM, the publicly traded person.”

This helps when I want it to apply concepts to the idea of KmikeyM, which is helpful when looking at economic principles, because ChatGPT is pretty interesting when you say, “here is something about x, now apply this to y.”

I’d love to know how you’re using it, what you think about it, and if you totally hate it (some people I know really, really, hate it and have good reasons).

Oh, and there is also the #disc-ai channel on the KmikeyM Slack if you want to talk about it more with some other shareholders.