Hacker News Reader

Comment by echelon | Hacker News Reader

I'm suggesting that ML will lower the barrier to entry so dramatically that the value of individual songs and musicians will plummet. The back catalog of copyright that the RIAA holds will become a fraction of its value today.

I could be wrong, but nascent technology in this field looks incredibly powerful.

I think this will happen with all media.

View on HN

"ML will lower the barrier to entry so dramatically that the value of individual songs and musicians will plummet"

It's hard for it to plummet down much further.

There's already a glut of content, and the vast majority of musicians can't make a living from their music.

The ones who do well tend to have giant marketing machines behind them to promote their music, and ML will do nothing to change that.

This... doesn't make any sense to me. Are you arguing that songs written by machine learning will become sufficiently good at their "job" that there will be no value in actual humans writing music? And that this will be a good thing?

Yes and yes.

People will still be driving and making money. But there will be more people doing it and catering to a much wider audience.

It'll flatten the curve. All long tail.

Patreon is the first hint at this.

You wouldn't even need ML. Iirc there was this guy who coded a script that played every permutation of 12 notes in a 5 minute period and then made all the Melodys public domain. Ianal so I don't know if this would hold up in court, but technically that should invalidate every copyright on Melodys after that point in time.

E: https://www.youtube.com/watch?v=sJtm0MoOgiU&t=135 found the ted talk

Since there is little to no creative input in merely enumerating permutations, I don't think he has a valid copyright to waive. But the fact that the data set exists might sway a judge or jury away from upholding copyright claims on trivial excerpts of other works.

Reply on news.ycombinator.com