If you’ve followed me at all over the past few years, you’ll know that I’ve been tinkering with AI for awhile now. In the early days of AI — like, five or six years ago — it was, similar to the early days of the Internet, basically text-based. You could ask it a question or give it a prompt, and it would write text for you, but there were no images, videos, etc., like there are today.
That itself shows you how far and how fast it’s come… but more on that later.
One of my own early experiments came in late 2019 with the then-current version of ChatGPT, which I believe was version 2.0 at that time. Having been a professional writer since my early 20s, I was curious as to how this new technology could actually write stuff from scratch. So I set up a copy of ChatGPT on my web host’s computer and tried a few things.
Most of the trials didn’t really get anywhere. The technology was young and still kind of stupid. Let’s just say “rough around the edges.” But one prompt I tried was something along the lines of, “Write a sermon about ‘Thou shalt not covet thy neighbor’s wife’.” I forget the exact prompt, or else I’d try it again now just to compare. But when ChatGPT spit out the sermon, I took it, created a little animation to go with it, and put it into a video. Perhaps you remember Brother Hal, The Artificially Intelligent Preacher?
You should watch this video before you go on, because it’s kind of necessary to appreciate what comes later. The sermon, written by the AI in 2019 and put into this video verbatim, is a good deal of nonsensical gibberish that sometimes almost sounds weirdly lucid, combined with a few unintentionally hilarious moments (“So, we should make sure that we are not engaging in any adultery, or adultery of the second degree”).
The point is, this was state-of-the-art artificial intelligence about five years ago. Let’s fast-forward to now.
AI Art
Images are obviously the next thing to come after text. A picture’s worth a thousand words, as we all know. So it was only a matter of time before AI learned to create pictures.
Here in 2024, AI can create pictures so realistic that you can’t tell them from actual pictures taken with a camera. It can create, visually, pretty much whatever you want: a cartoon, an abstract piece, an anime, just about anything. The image at the top of this post was created using AI and asking for “old style 1960s tv film computer flashing lights.” It took a couple of tries to get exactly the right prompt for what I was looking for, but there it is.
Here is a “cartoonish caricature of donald trump shrugging his shoulders with a confused look on his face.” The shoulders aren’t quite shrugging, but the rest is pretty good.
This image, and the one at the top of this post, were created at a site called nightcafe.studio. I use this site a lot. I’ve made the cover art for some of my music releases there. I’ve made images to illustrate various social media posts and suchlike. It’s a wonderful resource when you need some kind of image quickly and you’re not a graphic artist. Or if you just feel like creating something.
They have a free level and paid levels. You use “credits” to create images, and they give you five free credits a day, which can be good for up to 10 images, simply for clicking on a button. So you can do a lot without spending any money. Go try it. Type something into the prompt. See what comes out.
AI Music
As a musician, I’m also interested in using AI to create music. Pretty much ever since there’ve been computers, musicians have been using various algorithms combined with hardware or software synthesizers to create beats, melodies, rhythms, and so forth. Mostly you can’t use these “creations” straight out of the box, but instead you use them as rough building blocks that you then massage, combine, modify, and mix into something music-worthy. Musicians, composers, and producers do that all the time now. I’ve done it for a bunch of songs. The software gives you a start, and then you take that and run with it.
But that’s not AI. That’s just a pre-programmed algorithm. And even if the algorithm has a bunch of randomness programmed into it, so that it comes out different every time, it’s still not the same as the machine actually “thinking” and coming up with what’s basically a complete and complex original work from scratch.
Again, fast forward to now. I’ve tried out a bunch of music-making AI apps, and most of them — the free versions, at any rate — have been crap. They give you stuff that’s not particularly musical, and they play it on synthesizers that sound like they’re out of a 1980s “Pong” video game.
Then the other night I went to a site called musick.ai. I typed into the prompt, “smooth jazz with sax bass drums and piano“. That’s it, nothing more. This is what came out:
It actually created two different versions of the song. (The AI came up with the title as well.) This music sounds like it could be topping the smooth jazz charts nationwide. It has various instrumental breaks. It has rhythm and tempo changes. It has real-sounding instrumentation. It has… well, everything.
It took the computer about five minutes to come up with this.
Go there and try it yourself. You’ll be amazed. (You have to sign in with your Google account, and you can’t download the resulting work unless you pay. But otherwise, you can have at it.)
AI Talk Shows?
Anyway, no sooner had I been blown away by the compositional and performance chops of AI music than I came across something else that got my attention.
Here’s an article, published just last week, entitled “Google is using AI to make fake podcasts from your notes.” Basically, it says Google has taken its NotebookLM app — which is already armed with Google’s Gemini AI to help you analyze your notes and other documents — and beefed it up to use those notes create an artificial podcast, complete with two hosts who chat and banter back-and-forth about your selected topic of interest.
Say what? This I had to see.
So while I had an hour to kill the other day, I typed up a brief page about Donald Trump and whether or not he should be elected president. I listed three pros and three cons. I did this objectively, because I wanted to see the two hosts each take a side and make the arguments, sort of like a debate. But that’s not what happened.
First, here is the document I used to feed into the AI. Nothing more, nothing less. This is the only thing the AI had to go on.
Here is the resulting discussion it generated. Again, it took the computer maybe five or six minutes to do this. Just… listen….
Talk about blown away! I’m not really sure what to say about this. This entire discussion came from that single document that I wrote in about ten minutes and uploaded. It turned into this!
OK… Yes, I do know what to say. Listening to that discussion, I think the “artificial” intelligence is showing itself to be far more intelligent — and far wiser — than a lot of human beings with so-called “natural” or “inborn” intelligence. Maybe the future is not so bleak after all.
Let all this stuff sink in. It is really pretty amazing, and it gets moreso every day. Try out the tools. Be curious, and be amazed. And maybe create something cool.
More later.