Independent creators seem to prefer weaker intellectual property (IP) protections until it comes to AI. We can forgive a certain amount of hypocrisy here, since the original stance on IP probably wasn’t well considered, but it does make it hard to take these claims seriously. Saying “we should adopt policies that benefit me personally” doesn’t have the desired rhetorical heft, but few of the arguments against AI are any more convincing. A lot of this confusion seems to stem from the fact that we haven’t answered one simple question: what exactly is IP intended to protect?
If you think that IP exists to protect profits, then the opposition to AI seems more coherent. Haves (owners of IP) want stronger protections, while have nots want access to the thing of value. I won’t have to work as hard to market to my customers if I can just set it in the Assassin’s Creed universe and free ride on all the effort Ubisoft has put into making that name mean something. I will want free rein of the Assassin’s Creed IP, and Ubisoft is going to want to make sure it sees the rewards of its marketing efforts. IP becomes a simple debate about the priority we give to the haves and have nots. The problem with this reasoning is that IP is not about protecting profit.
IP law is enshrined in the US Constitution. The US Congress is empowered “To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.” Unlike, say, the US Postal Service, the clause includes both the power and the reasoning for it. IP protections exist to promote ideas, not protect profits. The exclusive right certainly is intended to let creators benefit from their creations in the form of profit, but that is the mechanism through which the real intent — preserving and promoting an engine of progress — is realized. It is why we get justifiably angry with something like the Mickey Mouse Copyright Act (named both for its chief proponent, as well as the quality of the legislation), which serves only to enrich existing owners of IP at the expense of delaying society’s broader benefit from those ideas.
I, like the vast majority of the world, am not bound by the US Constitution, but that document is going to have a heavy influence on the rules AI will be governed by. The protection of ideas instead of profits also seems like a reasonable goal to aim for and provides a certain clarity to an otherwise murky debate. Owners of IP are only assured their profits if they continue to buy or generate new creations. Creators are only assured the opportunity to profit from creations that people find sufficiently useful or entertaining.
Adopting the perspective of protecting ideas instead of profits explains why it is so difficult to copyright so many parts of entertainment. Ideas cannot be copyrighted, but implementations can be. Creators are free to make “Doom clones” until we call them First Person Shooters, but they cannot have Cacodemons and they cannot use the source code without Id Software’s permission. This system is not necessarily fair! Innovations in gameplay can involve a lot of time and effort to develop while also being easy to copy. It is, however, preferable to a thicket of rights where creators are paralyzed by the possibility that what they’re working on bears some similarity to a post on Tumblr that will later claim it was the original idea and demand the revenue.
This perspective can also shed light on why some of the arguments against AI are so unsatisfying. It may be common to argue that AI generated products are bad because they don’t require a lot of effort, but effort is only loosely related to quality. An author who bangs out a pristine first draft is as entitled to copyright protection as an author whose work ethic exceeds their talent. It is the audience that will decide which of those exclusive rights receives the greater reward. Likewise, protections are not based on the quality of a work. Ed Wood gets the same treatment as Joseph Heller.
The constitution also extends those rights to the author, not the tools. This is useful for understanding the reasoning behind the US Copyright Office’s report on the copyrightability of AI outputs. The whole report is worth reading, but, in short, prompts are not copyrightable (no matter how much effort you put in), and the outputs of AI are not copyrightable (no matter how much curation you perform), but assembling the outputs of AI into a copyrightable work is. Crayola can’t claim your drawing, and it can’t register the coloured pattern that randomly forms on the top of the packaging as it’s opened and closed, but you still retain the exclusive right to whatever you draw with their product. Copyright isn’t just about preserving an engine of progress, it places humans at the center of the process.
Copyright is not the only lens to view this issue through, but it is notable how an application of the existing rules produces a fairly common sense understanding. It is also fair to say that the major beneficiary of the current IP regime is the consumer. This perspective promotes humans doing the work, but the ‘doing the work’ part gets just as much emphasis on the ‘human’ part. Doing the work is very difficult, and it becomes harder to compete as more people are enabled to create things. But the fact remains that the protection exists so that we can continue to do the work.
SystemTalk 