Suppose that someone has created an advanced AI platform that allows the creation of original music in the style and voice of a particular artist. The copyright law here is clearly unsettled.

Now, consider the following quantitative example. A legislature is deciding how much of the royalties for such a song should be given to the imitated artist. This can be between 0% and 100% in increments of 10%. Suppose there are 5 artists who want 100% of the royalties, 10 AI creators who want the artist to receive 0% of the royalties, and 3 platforms who want a 70-30 split between artists and AI creators. Assume the loss around this "ideal point" for each type is given by |p-x|, where p is the ideal point and x is the realized policy.

Now, suppose instead of trying to pass a law, the AI creators decide to lobby regulators for changes in enforcement of copyright. The creators want to argue that giving all the royalties to artists will eliminate the incentive for the creation of AI generated art. Suppose the probability of success is a function of the money invested in lobbying. Specifically, the probability of success equals 2 !, where x is the total lobbying expenditure in millions of dollars. The value of a successful policy change will be $5 million, and the creators are risk neutral.

Delving deeper into the lobbying process now, consider the regulator. He is concerned that leaving the royalties 100% to artists and 0% to creators may eliminate incentives for creators to generate desirable work. Suppose he perceives there is a 1/3 chance that AI creators are driven by intrinsic motivation so that they will work just as hard without the chance for royalties as they do when there is a chance for royalties. Suppose he also knows that the lobbyist is honest with 50% probability and will just say whatever his client wants him to with 50% probability.

1) Under the same benefits as above($5million, regardless of whether the policy is beneficial to creativity incentives or not), how much should the industry as a whole (assuming they can coordinate) pay the lobbyist, assuming that they also believe that the lobbyist will be honest with 50% probability, if they know they are intrinsically motivated, and assuming the regulator will only implement the policy if he thinks there is at least a 50% probability that the policy is beneficial?

2) What if they know they are motivated only by royalties?

 

3) There is something weird and somewhat inconsistent about this problem. What is it?

Answer rating: 100% (QA)

Analyzing the AI Music Copyright Conundrum 1 Lobbying with Intrinsically Motivated Creators Expected Benefit 5 million if lobbying succeeds Cost Lobby... View the full answer