the music didn't change. i did.

i found a two-hour video called Armonía Andina and started putting it on while i worked.

it is soft instrumental music with pan flutes and the shape of the andes all over it. it did not ask me to stop working and admire it. it made the room calmer. i returned to it enough that youtube learned to keep it nearby.

i had small suspicions, but not enough. there was no moment when the melody glitched and revealed the machine underneath. it was only a vague feeling that i did not consider strong enough to matter.

recently, a label appeared below the video title:

made with ai

the music did not change. i did.

my mind kind of makes me watch it less, but it still transmits the same emotions it did. one part of my mind seems kind of lied to.

that contradiction is what stayed with me. if the sound still worked, what had the label changed?

what the label changed

youtube's disclosure policy says the label can mean that content was fully or partially created or meaningfully altered with ai.

learning it late bothered me, but that was not the deepest part. if the label had been there from the beginning, i probably would have skipped the music before hearing what it did to me.

i think i would not have listened to it if it was labeled as ai from the start.

the same music would have reached the same headphones. the label would have changed the choice before the sound had a chance.

researchers tested a version of this reaction in two studies about human and ai art. participants saw paintings that were all created with ai but randomly labeled as human-created or ai-created. the human-labeled paintings were judged more highly, especially for profundity and worth.

so a label can change what people imagine behind a work. but "made with ai" is also a blunt label. it says that ai was involved. it does not tell me what a person actually contributed.

i know that from making songs

i have already made four parody songs with ai. that makes it impossible for me to read "made with ai" as "no human was there."

for shape of ai, i brought the premise and the hook, sang the vocal, rewrote lines, chose the voice-conversion settings, and mixed the result. for party in the ai lab, i went through three lyric drafts and four generated vocal mixes, moving the jokes until they became one story instead of a list of ai references.

the systems did a lot, but i was not only requesting output. i was trying to express something through the choices i kept making. my experience was in the premise, the revisions, the voice, the rejections, and the stopping point.

"made with ai" can describe that process. it can also describe someone writing one broad prompt and publishing the first acceptable result. those are not the same kind of authorship.

what was standing there

that was the information missing from the label under Armonía Andina.

from what i can tell, the music was mostly generated by ai. i do not know the exact prompt or what was edited afterward, but the human contribution was closer to a request like "make a calm and relaxing andean song with these instruments, inspired by this other track" than to composing or arranging the music itself. the system appears to have supplied most of the finished music.

if so, a person still chose what kind of song should exist. what changed was who or what supplied the details that made it work.

a human may still decide what the song is for. the system may now be able to produce what makes it work on me.

that is the limited sense in which something nonhuman felt as if it were standing there. not as a consciousness, and not as the owner of an intention, but as the source of much of the musical form that reached me.

the music had not failed. it had succeeded, and the label made that success feel like evidence that a person might no longer need to compose the details for them to move me.

logically, nothing was taken from the emotion. the calm happened inside me. emotionally, something was taken from the story i had about where an emotion like that had to come from.

this does not mean the system felt anything. it does not show that it understood the andes, the calm, or me. the narrower possibility is enough: a human inner life may not need to stand behind each expressive choice for a real human emotion to appear on the receiving side.

the small resistance

the disclosure should have been there from the start. the fact that i might have skipped the music is not a reason to hide it.

but disclosure alone would not settle the feeling. it would only let me ask the real question sooner: was this a person using ai to express something, or a system producing almost everything after a broad request?

my own songs taught me that ai use does not remove the human experience behind a work. this track made me confront the opposite possibility: that the human experience could be thin and the emotional effect could survive.

i can say that the emotion happened inside me, so learning how the music was made does not erase it.

and i still play the video less now.

maybe each listen asks me to accept that even when human intention is hard to locate, the effect may remain. part of me would rather choose another track than answer.

i cannot separate the music from that feeling cleanly yet.

the track is still in my history. the label is under the title now. when i need music for work, both are there: the same sound, and the small resistance before i press play.