Raw LLM Responses

The nature of music is very modular. You take these very small phrases and combine them to form melodies. The more coherent the music is the more it's going to overlap with other music because the number of modules you combine is getting smaller the more "likable" the music is. I can't remember most vocal trance songs. The melodies are writen over the same chord progressions and are very similar. So what makes things different there. The way an individual sing in their particular personal way. The rhythmic pattern and specific word combinations. There is a lot going into a song that is performance. And performance alone can't be just written down. Filter movements can't be just writen down. The very unique way someone uses articulation can't be writen down. Why am I saying that. You give the task to ten musicians to record a song based on sequence of notes. Let's say A B C. Just that Repeat A B C. The only requirement is that you keep repeating A B C. All ten musicians will come up with ten diferent compositions. What does that have to do with AI. AI has stored songs from different artist. In some cases it spits out entire melodies that are already copyright protected. Most times it has set of phrases it's pulling of this material and trying to combine them. I have noticed some repeating patterns AI uses. It's very possible many people to get nearly identical songs all the time. No it's unavoidable. You can create something generic without AI but you can say I wrote this in my shower...blah, blah. But when you prompt AI people are going to look at it differently because it wasn't created in the shower. Wait this is getting too long. Bottom line. Performance is really what copyright has to emphasize even more than melody because we are running out of unique melodies that are also appealing to the masses. That's why today's pop hit may have the same melody as a song written 60 years ago. Maybe isn't always stealing. Maybe they just want it to be melodic.