If it seems like I’m writing about artificial intelligence a lot lately, that’s because every other day there’s yet another astonishing and dreadful advancement. I’m just trying to wrap my head around this stuff.
Okay, make sure your volume is turned up, because what you’re about to hear is amazing. And funny. And, if you value the American experiment, terrifying. But AI is here and it’s not going away, so we need to be prepared for it.
The last time I paid any attention to AI “music,” the results were so bizarre and unnerving that I wanted to crush every computer in the world like Sarah Connor at the climax of The Terminator. It hurt my brain, and my feelings.
So when I heard about a new music AI called Suno, I was skeptical. How good could this thing be?
I first read about this “ChatGPT for music” in Rolling Stone, which is still a pretty good news outlet from time to time. Author Brian Hiatt went to the Suno offices in Cambridge, Massachusetts, and they showed him how the AI reacted to a text prompt: solo acoustic Mississippi Delta blues about a sad AI
.
Hang onto your hat, human.
A robot made that. There’s no actual guitar. Nobody wrote or sang those lyrics. It’s all synthesized algorithmically.1
And when I listen to it, part of me believes a robot has the blues.
So of course, then I did what I always do when a new public AI intrigues me: I challenged it with a text prompt containing the least likely combination of words I could manage. Just the dumbest, most random crap.
I tried to play Stump the AI Band. “Oh, you’ll crank out a song about anything, you dopey toaster? Well, how about this?”
And as you’re about to hear, I failed utterly.
Here’s what the robot did when I typed in A cookie-monster
2 death metal song about playing kickball at school
.