When AI Sings the Blues

“I hate that I love this.”

That was a message I received from a friend. The next text was a link to a song. We both like similar kinds of music, so we often send each other tracks we’ve discovered. I hit play without reading the details, assuming it would be a good song ruined by bad lyrics or something along those lines.

It was a beautiful soul song with a vintage feel and a singer with a voice like Otis Redding. I started to recognize the lyrics. It was a remake of an old punk song we both like. Did she hate loving it because it was a remake? Who could be mad at a cover this good?

I looked to see who was performing the cover and realized it was AI-generated.

Oh. I get it now. An AI-generated soul song does not feel very punk. Honestly, I was a little disappointed in myself that I, a soul music lover, could not tell that beautiful voice was more Bot-is Redding than Otis Redding.

 

We are not alone in our dismay.

With the rise of AI-generated music on Spotify came a flood of user complaints. In response, the platform rolled out a feature in late April that verifies artists as human. How does someone verify they are human? Spotify looks for things like concert dates, merchandise and linked social media profiles. AI-generated personas and artists cannot receive the “Verified by Spotify” badge.

I can understand why people want to know. There is a sense of deception when you discover that a piece of art you are enjoying is made by AI. Once you know, having an emotional connection to the experience that made the art beautiful suddenly feels absurd. It takes lived pain to sing like Otis Redding. You connect to those emotions when you hear him sing. You are not sure what you just connected to when you are moved by bots cosplaying emotions they can never experience.

That is not to say artists are not using AI in brilliant ways, but clarity around where the artist ends and the AI begins is essential to framing the experience. The soul song was beautiful, but I would not consider it a work of art. Anyone can ask an AI to generate a vintage soul version of a punk song with a singer who sounds like Otis Redding and receive essentially the same result. If the emotional contribution is coming from the AI, it is more likely to make people uncomfortable than to move or convince them.

 

“Inspired by Real Frustration”

I had a similar experience recently when I saw a commercial for the streaming service Frndly TV. At the bottom of the screen was a small disclaimer that read, “AI actors inspired by real frustration.”

While I appreciated that Frndly disclosed its use of AI actors, despite one actor’s unnaturally wide smile already giving it away, the disclaimer did not make their frustrations feel more believable or less absurd — not because the AI actors were bad performers, but because I inherently knew they had never experienced frustration.

If I explain a frustrating experience to a human customer service agent and they tell me they understand, it can feel comforting. The same line coming from an AI agent feels hollow or even insulting. An AI actor describing an exhibition’s themes in an informational video may feel reasonably educational. An AI actor wiping away a tear while describing how an exhibition moved them feels inappropriate or even comical.

As the use of AI agents, actors and artistic outputs continues to grow, we will see this line tested more and more as we determine which aspects of humanity we are comfortable accepting in digital form. And, like Spotify, we will continue to see an increase in labels and badges like “verified human” or “made by a human.” After all, can music without a soul ever truly be soul music?