Episode Details

Following on from last week’s piece about the extent to which I use AI, I’ve had a surprising number of messages asking which AI I actually use and what for.

I should immediately stress that I am not some sort of AI guru. I know people use Claude to write code, automate businesses and build entire internal operating systems. That is beyond me. I can’t code. I’m a one-man band, who occasionally hires freelancers. I’m self-taught. But here’s what I actually use and what for.

I stress the best method of all is trial and error. You get results quickly. If you don’t get what you’re looking for, adjust the prompt, or try a different app.

Let’s start with the visual stuff.

Pretty much every image accompanying my articles, such as the one above, is generated on Midjourney. I’ve experimented with ChatGPT, Grok and other image generators, but I like Midjourney’s images the most.

My prompt is often just the article title plus the aspect ratio. Four options appear. I pick the best one.

That alone would have seemed miraculous ten years ago.

I also use Midjourney extensively for music videos. For example, in this video about the lighter side of hyperinflationary collapse, almost every visual was AI-generated from the lyrics. My editor, Goat, then used Runway to animate the images. We filmed my face against a green screen and plonked it on top afterwards.

If Midjourney didn’t produce what I had in mind, I simply kept adjusting the prompt until it did, or I tried another image generator as a last resort

Even as recently as five years ago, let alone twenty, to make a video like this would have cost hundreds of thousands, millions even, and taken many months. We would have needed teams of animators, post production specialists, Soho studio space and lord knows what else. That, to my mind, is the genuinely revolutionary part of AI. Democratisation of media and all of that.

But on top of it all you still need someone - in this case my editor Goat - who knows what they’re doing.

People often argue that AI ia replacing creativity. What it is actually doing, at least in my case, is dramatically lowering the cost of production and making creativity available to all. The possibilities for creative littlemen like me are enormous.

I made this video using Grok and Neural Frames

And this one was generated entirely in Neural Frames

Ironically, we used no AI in the music itself. We edited the videos either in Capcut or FinalCut.

By the way, if you enjoy these videos, the first place I upload them is at my comedy Substack, so sign up to that. It’s free.

Writing, research, advice and more

This next video, about the most prolific slaving civilisations in history, generated millions of views across social media, and became the most viewed page on this Substack. It is an interesting case.

Not because of the images themselves, which were generated with Midjourney, but because of the research. AI couldn’t and in some cases wouldn’t do it.

Claude flat out refused because of the subject matter. It would not engage. (IN other words it is biased). ChatGPT couldn’t get its head round what I was trying to do. Grok came closest but in the end I worked with a human researcher, Sam, who I knew from my book, who turned out to be much better.

I have paid subscriptions to Claude, ChatGPT, Grok and Venice.

I cooled somewhat on Claude after the slavery episode. Around the same time I was