Is the catch that the entire song was AI-made (with perhaps some human guidance)? I kind of got that vibe.

@Calibrate Yes. You can see below about that.

Almost all the "no" voters didn't actually listen to the song based on the webserver stats. Given others I've talked to, I suspect that out of the people who listened to it, about half of them liked it, but the biggest complaint was compression artifacts rather than musical composition.

I think that the AI tools can get pretty close to indistinguishable from the "old method" of writing music, but in this particular case, I know that I didn't use the prompts correctly.

I confess, I hated it ten seconds in and gave up after thirty seconds.

If you voted, then you didn't follow the directions.

But whether people like the song or not, while interesting, wasn't the point of this question. The point was what the accuracy of poll results on Manifold can be.

I am tracking statistics about the music. In this case, at the time of your comment, there were four votes, but only you actually listened to the song. That means that I (and you) are mostly talking to bots, and it confirms my fears from the awful discussions I was having in other markets.

@SteveSokolowski does it track full listens only?

@CraigDemel Yes.

I also spent three more hours on the song this morning and significantly improved it and changed the link in the description, so it will be interesting to see what the ratio of full listens to "likes" is starting now. The improvement was required starting at the bridge, so my guess is that nobody will actually listen to the whole thing and it will change nothing about the poll.