The Writer’s Strike is over, and it looks like it was a big win for the writers. But much of the narrative was about AI and its potential impact on writers’ jobs. So what did the agreement say about all of that?
Well, not much. Here is all the 100-odd page agreement has on the subject.
Let’s see if I can summarise:
Generative AI is a thing, and we agree that we pretty much know it when we see it.
Generative AI is not a writer because it isn’t a person and, therefore, cannot produce literary material.
If you give a writer some AI-generated output, they can treat it in the same way as if it were a blank sheet of toilet paper for the purposes of determining their compensation. If the writer uses AI-generated output or tools, that will be ignored in determining their compensation.
A writer must ask permission to use AI in their job. (Basically, we need to get legal to sign off that this won’t be a copyright issue.)
A writer cannot be forced to use ChatGPT to produce literary material.
We agree to be able to sue each other for stuff not really covered in these couple of pages so that the lawyers, at least, are not out of a job.
We agree to get together to talk about this stuff regularly because no one really knows where this is all going.
I’m no lawyer, but that is what popped out to me. If I translate this into economics, this is what I get.
Machines and the owners of machines aren’t getting property rights on any of this stuff.
If any AI is used in the writing process, that is on the studios and not the writers, who will get their paychecks as if the AI never existed.
Both studios and writers can veto the use of AI.
Interestingly, none of this precludes a great expansion in the use of AI in screenwriting, but it just says that everyone has to be happy with it. From a financial perspective, writers are protected, so that won’t be a reason they aren’t happy. It will be more of a ‘creative thing’ or something intangible that may cause some disagreement.
On the flip side, if the studios think the writers are now phoning it in by using AI, they can tell them not to.
In other words, they can confine any arguments about the use of AI to the use of AI and not necessarily need it to extend to some broader disagreement that might lead to a strike or other forms of disruption. In other words, the more AI is mutually beneficial, the more it will be used. That’s a hallmark of a reasonable degree of efficiency.
That said, this is quite interesting. One of the things that is often missing from debates about the use of technologies that can impact jobs is one Ronald Coase — that if there are gains from trade, there should be a deal that can be made to make that happen. What prevents that, in reality, is that one side has asymmetric power and decides to use it to get all the gains themselves rather than come to an agreement that shares the spoils around. One suspects that bloodshed could have been avoided if that had been possible in the time of the Luddites. (See a great recent Cautionary Tale from Tim Harford for more on that.)
As usual, the writers recognised what many others have is that “AI sells a story”, and so that is what happened here. The strike was really about other stuff all along.