The other day, while driving, I put on a playlist I hadn't listened to in a long time, and an older song by Erlend Øye came on. I was enjoying how distinctive it felt when I noticed that part of my pleasure came from something quite separate from the music: I felt certain that no generative AI had been involved in making it. That certainty had become part of the experience.
That's a peculiar thing to appreciate about a song, because you can't hear the absence of a tool. But apparently you can feel relief at believing it was absent. Somewhere along the way, being able to enjoy a piece of music without wondering how it had been produced had become a quality of its own.
Then I thought about the machinery behind popular music, where songwriters, producers, performers, and record labels have shared the work for decades. The name on the cover has never necessarily belonged to the person who wrote the words, and knowing that has rarely stopped us from finding something personal in a recording.
Suppose Dua Lipa sings a song with lyrics written by ChatGPT. Is that fundamentally different from singing lyrics written by a team of professional songwriters, and would the answer change if the model supplied one phrase rather than the entire song? Her interpretation could still be expressive; choosing, revising, and performing would still involve creative decisions. I don't think counting the participants gives us a simple measure of artistic worth.
But even if I could answer all those questions, I'm not sure it would resolve what I had felt while listening.
For me, something began to change around 2024 or 2025, although I can't locate a precise moment. But from 2026 on, I increasingly started carrying the same question into encounters with songs, illustrations, photographs, and beautifully written paragraphs: how much of this was generated?
People will keep making things without generative AI, and sometimes we'll know exactly how they made them, but that knowledge has become something we might need to establish rather than simply assume.
Of course, every musician has listened to other musicians, and every writer carries other writers into the room. Everything is a remix, as we keep reminding each other (h/t Kirby Ferguson). Yet influence never stopped me from imagining another person working through what they had inherited. Generation introduces a different uncertainty about what the person actually contributed, and that uncertainty can follow even someone who has never used these tools.
I keep thinking about how a washing machine appeared on both sides of this transition.
In the 2017 documentary AlphaGo, Fei-Fei Li says, "We're really closer to a smart washing machine than Terminator." She's describing the limited capabilities of AI at the time, using an ordinary appliance to bring the conversation back from science fiction. If you haven't watch AlphaGo here
Seven years later, on March 29, 2024, the writer Joanna Maciejewska gives the appliance a different role: "I want AI to do my laundry and dishes so that I can do art and writing." She objects to automation taking over activities she wants to keep while leaving the unwanted work in place.
You know what the biggest problem with pushing all-things-AI is? Wrong direction.
— Joanna Maciejewska — SFF Author (@AuthorJMac) March 29, 2024
I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.
Fei-Fei was talking about capabilities, while Maciejewska was questioning priorities, so their remarks don't contradict each other. What interests me is how the washing machine moves from reassuring us about AI's limitations to reminding us what we hoped it would do. A system could remain limited in many ways and still change our relationship with creative work; it didn't need to become conscious before I started wondering whether another person had written the song.
The question becomes stranger when we consider who the work is intended for. The Dead Internet Theory imagines (or chronicles, as things seem today) an internet increasingly occupied by automated content and automated responses. An internet by agents, for agents.
An agent could write a song, another could review it, and a third could recommend it to other agents, producing the appearance of cultural activity without anyone stopping to listen. Some exchanges between agents would be useful, and I might welcome software that reads or sorts things on my behalf. Still, there's something unsettling about imagining creative output whose production and reception could proceed with human attention optional at both ends.
The uncomfortable part is that I'm writing this with AI, and I understand why someone else would choose to do the same. The playlist and what I felt while listening to it are mine, but I'm working with a model to find the words for that experience, which means the uncertainty I now bring to other people's work also belongs to these paragraphs. I find the help useful, even as I miss the confidence with which I used to encounter something another person had made.
I occupy both sides of the change, using the tools and helping normalize their presence while discovering, as a listener and reader, that something about the experience has shifted. I can admire what they make possible without pretending that every consequence feels like a gain.
Listening to that song, I missed a kind of trust I had never thought to name. In contrast, there isn't a single song that came out after 2023 that I can confidently assure is AI-free. But not so long ago, I could imagine someone making something for somebody else to hear, and that day I was that somebody, enjoying the encounter without having to investigate the machinery behind it.
Part of the magic was never having to ask.
PS: In case you were wondering, this was the epiphanic song: