Kevin Surace generated a full four-minute song with AI vocals in three minutes, impressing music professors so much half wanted to kill the technology and half wanted to use it. He saved $10,000 in studio costs and sees the split reaction as proof the AI actually works.
This AI expert created a song in minutes, and the musician's reaction was not what you'd expect. I speak around the world on AI and the impact of AI. I was speaking to a group of music professors about what we can do with AI and music. And a few years ago, when we started getting AI that could start to write music, it was pretty crappy, but interesting and fun, right? Now you can actually write amazing songs, have demo versions of them, including people singing them. Like, it'll save you $10,000 of studio and production time, and you can have it in 5 minutes. So I took them through it, and in 3 minutes I was done. Then I said, listen to the song, it's 4 minutes long. We listened to the whole thing, and they're listening intently. And this time, because it's so good, it's indistinguishable from real. Like, you think this is a Broadway singer singing this song that didn't exist 5 minutes ago, 4 minutes ago. So they listen to the whole thing, and I've got people screaming at me, it's crap, it's junk, it's terrible, it should never see the light of day, it's not fair, it's this, it's that. That's when I knew it was good. Like, before they'd say— The thing's junk, no one's gonna listen to it. Yeah, and they'd write it off. Now, they were mad, right? Some of the people, not all of them. Half said, how can I use it? The other half said, I will never use it in my life. It should be killed. And I go, huh. So it must be working really well when you get that kind of visceral reaction. So when you have coders today saying, I will never use that, it writes junk code. Oh, you're jealous. Okay. Jealousy is not a pathway to success ever.
