French music streaming service Deezer says around 50,000 completely AI-generated songs are uploaded to its platform every day. While many of these songs don't reach a wide audience, a few have been listened to millions of times over the past year.
This raises the following questions: What would our future look like if it were filled with this kind of AI music?
Denis Bechard is a senior science writer at Scientific American.. For the better part of a month, Bechar only allowed herself to listen to her own AI-generated music using the AI music app Suno. He says the experiment is an attempt to think more critically about how we interact with this type of music in the future.
Mr Bechar said: Today's explanation Host Noel King talked about what he's learned so far and how his AI work compares to human-made music. The conversation has been edited for length and clarity.
There's more in the entire podcast, including excerpts from Bechar's songs, so be sure to listen. Today's explanation Get your podcasts wherever you want, including Apple Podcasts, Pandora, and Spotify.
Well, I use Suno to create songs.
When you come up with prompts and plug them in, each prompt creates two songs and you try to be as creative as possible. I usually plug this in two or three times and vary it, adding different types of instruments and different types of vocals, and plugging them in a ton. The song that made me laugh was called “Organ Trafficking.” I had requested a contemporary rap song with female vocals, and playful and sarcastic lyrics. And so this song was born, which is like a metaphor that revolves around organ trafficking. I was quite surprised.
One thing I've noticed is that a lot of the mainstream music I listen to is heavily processed music, music that's designed to capture a large market. It's not very personal to me anyway, so in that particular context, [the music I made with AI] In many cases, I didn't notice much of a difference.
If someone hands you a playlist of 10 songs, 5 of them are AI and 5 are not. Do you think you can tell the difference?
oh. What does that tell us?
That means the AI is getting a lot better.
One of the things I realized during this process was that a lot of popular AI music is being listened to by people on Spotify, which has millions of listeners. [are] A very soulful, very gritty song.
It's like “Don't Tread on Me” by Zania Monet or Solomon Ray or Kane Walker, and Kane Walker is not human. It's an AI avatar, right? Or “Livin' on Borrowed Time” by Breaking Rust. All of these songs feel really authentic. This person has truly experienced these things and felt these things. That's how they meet.
I think AI tends to be most effective when it leans into its believability. Because AI can help us overcome the cognitive dissonance we think. This song isn't really a deeply emotional song, it's a departure from mainstream music made by humans, music made by humans. Music created by humans is often very heavily designed to be a summer hit or go viral in some way. And a lot of times you don't have that level of authenticity, that sense of authenticity. When AI reproduces it, I think we'll find it even more superficial or artificial because there's already an artificial element to it.
Do you plan to continue making AI music once the experiment is over?
Oh my god, you love power.
What struck me about this song is that you'd walk somewhere and think, “What would happen if I asked you to combine these styles or put a banjo on a hip-hop track and add this kind of vocals? What would you get?” Now I'm interested.
I can say that now I don't have to worry about connecting with people. At first it was. At first, I was really like, “Who is this person?” When you're halfway through a book and think to yourself, “What kind of human thinking led to this book?” You flip the book over to see who the author is, Google it, and you're like, “How the hell did they come up with this?”
In the beginning, I often felt the urge to know who felt this, who thought this. It just creates cognitive dissonance. I said, “This is a machine. This machine didn't fall in love. This machine didn't have these experiences. This machine didn't wake up at 2 a.m. and write this song to express itself.” Actually, I was very worried. It kind of prevents me from enjoying the song.
So I thought. “Wouldn't it be easier if someone created an AI avatar and gave it a personality, and it was a fictional character that existed in the Metaverse, and that AI avatar was a song maker and was singing this song?” And strangely enough, it does. That will make things a little easier. So I was just imagining these AI avatars and I was like, “Okay, I'm imagining a fictional character singing this song.” It lasted about 4 or 5 days, but eventually I got used to listening to music and stopped thinking about it.
Has doing this experiment and seeing how you react to this music changed the way you think about AI?
My conclusion from this is that in 10 or 15 or 20 years, there will be a lot of teenagers looking at the discussions we're having now and thinking, “What are these people saying? This is completely normal. Why are some people so conflicted about this?”
I think I can adapt pretty quickly. That's my intuition. There are a lot of big questions about protecting creators and artists, and what it means to be an artist. This raises a lot of questions, and I really want artists to be protected as much as possible and compensated appropriately. However, I think this will integrate into our lives much more smoothly than we currently think.
