How I've Used ChatGPT & Midjourney. They're fun & useful.
The current crop of AI products is fun and useful, though not dissimilar from what people can do. This will change over time, but I think we're off to a good start.
I turned this post into a short video, if you prefer that. I mostly talk about Midjourney in the video.
The influx and output of generative AI products makes me think their current superpower is speed & cost. GenAI has ingested billions of points of data, whether text, audio, video, images, or something else, and uses that to quickly create new content.
I think of GenAI like The Flash from DC Comics: ridiculously fast, but the ends (for the most part) are no different from what we can do. For example, if the Flash wants Naples-style pizza from Naples, he can get it in a second. For us, it would take longer and be more expensive, but the result would be the same.
With some caveats, it's the same with GenAI. I've used ChatGPT & Midjourney for hundreds of questions and images. So far, the quality of the result isn't dissimilar from what a person could come up with, depending on their familiarity with the topic. But the speed and cost of the response are unlike anything a person could come up with.
I want to publish another book. I'm in the process of publishing my first, My Fellow Americans, without an agent but it's advisable to have one. You usually reach out to, or query, agents yourself. To do that, you need some agents to reach out to. Sure, I could Google them. But when it comes to simple facts like this, ChatGPT shines. I asked it to provide a list of agents that have some history authors as their clients, and it came up with a list of a few dozen, complete with name and agency. I still have to reach out myself. But now I have a list to work from. Best of all, I didn't have to scrounge around different parts of the internet to compile it. It came to me, in a manner of speaking. The result is no different from what I could've come up with by Googling or reading the Acknowledgements section of a book, though far quicker.
I did some validation after the fact and most are real agents. Whether or not they're accepting queries now is another matter and not one ChatGPT can help me with, but it's a good start.
The cost was zero and the time to find them was minutes, not hours or days.
Another interesting use case is using ChatGPT to write short scripts. or one-liners. Here, I'm referring to a short program that does the same thing reliably.
ffmpeg is a useful and notoriously difficult-to-remember toolkit that's often used for audio & video tasks. Conversion, encoding, transcoding, etc. – whatever you want to do to an audio or video file,
ffmpeg can do it – if you can remember the incantation. But using plain English, you can tell ChatGPT what you want
ffmpeg to do, and not only will ChatGPT tell you what the commands are, it'll explain them to you.
I started using Midjourney in September to generate images based on lyrics from songs that I would then stitch together to create a lyric music video. If I wanted to commission the art for these videos from individual artists – which could be little as one image for a single-lyric song or dozens for something like Black No. 1 by Type O Negative – it could cost hundreds or thousands of dollars. As important, it could take days, weeks, or months to have all the images ready.
That's not the case with Midjourney. For $30 a month, I could ask it to create an unlimited amount of images, in almost any conceivable style. It would take a few seconds to generate each picture and I would save a lot of time and money. It's cheaper and quicker than any artist, though the result isn't as good (though this is a matter of taste). Still, Midjourney is improving daily.
It could be useful for artists as well. They can feed their clients commissions into Midjourney and have it provide something they can work off of. They can come up with something quickly, then make further edits in Photoshop. Artists can now produce content even quicker, and no longer need to start with a blank canvas unless they want to. Turnaround times can improve, and they can take on more clients.
As much as I like most of what Midjourney outputted, I can't say I was completely happy with all the results They were all somewhere along the way to what I wanted, but they took enough detours that the result wasn't precisely what I desired. For the time & cost that went into the quality I got, I'm satisfied. But right now, it's a case of something is better than nothing. It's enough to pique my interest to continue using it. I think it's a great tool to use as a starting point. If you have an idea of what you want, you can use Midjourney to come up with a quick and cheap prototype. You can take the prototype to a designer or artist to get something pixel-perfect, or more in line with what you want.
I was listening to episode 504 of ATP a few months ago. One of the hosts, Marco Arment, was/is developing an app about the tide levels in his area. The app needs an icon. Marco used DALL-E, OpenAI's image generator, to come up with some ideas. This is the bad one and this is the good one. As he mentioned in the episode at around the 16:35 mark, it gave him a jumping-off point – something to take to a professional designer who could then create a better final product. But it helps clarify communication. He can say "I want something like this, but pixel perfect" or "Something like this, but better than this." Then he can give direct feedback that the artist can incorporate – something Midjourney is incapable of. He can say "I like this and this, but I want that to connect to that over there."
When it comes to visual mediums, showing is better than telling but you still need to be able to explain what you want. Right now, you can do that with a human but not with a machine. The machines need to get better.
These products will continue to improve. But the genie is out of the bottle. Pandora's box has opened. What's done is done. There's no going back to how things were before. GenAI will be what we make of it. It's in our best interests to try and understand this new landscape and use it to our advantage instead of shying away from it.