Firstly, there's an enormous corpus of music out there that precedes this effort, and those copyright holders can still make claims.
Secondly, and perhaps more importantly, how would someone prove that a musician copied the work from this algorithmically generated collection? These haven't been number one hits or in widely available collections. Someone could not conceivably argue that a musician derived their work from this due to familiarity with the melody.
I wish the copyright problem would go away. But I really don't see this as a solution.
A large number of musicians and software designers need to collaborate on a project that is much more permissive with regard to sampling, covers, arrangements, adaptions, etc, all the while preserving the rights of songwriters who just happen to write a number one hit song to get paid for their work.
The whole thing needs to be lucrative enough that composers, performers and the tech companies involved can afford to flip the middle finger to anyone exhibiting corporate greed.