The Entertainment Industry Keeps Doing Our Homework

Someone has done the research again. Detailed, practical, brutal — the kind of document that makes you feel slightly embarrassed about whatever half-baked framework you were using last week. Geared toward the entertainment industry. Applicable everywhere.

This keeps happening.

It makes a certain kind of sense if you think about it, which I have been doing for longer than I'd like to admit. The entertainment industry spent most of 2023 on strike — the writers, the actors — and a significant portion of those negotiations were explicitly about AI. Not in the abstract "what does the future hold" way that tech conferences discuss it, but in the very concrete "are you going to replace me with a synthetic version of myself" way. Those are different conversations. The second one produces better research.

When your livelihood is directly and immediately threatened by a technology, you don't write a think piece. You hire lawyers, you bring in technologists, you commission studies, you read everything, and eventually someone produces a document that is — to use the technical term — actually useful. Not because the entertainment industry is smarter than everyone else, but because they had something specific to protect and a deadline to protect it by.

The tech industry produces a lot of AI research. Anthropic, OpenAI, Google, the whole apparatus — constant output, incredible volume. But most of it is either capabilities research (look what we can do now) or safety research in the philosophical sense (here are frameworks for thinking about alignment). What the entertainment industry produces, under duress, is deployment research — what happens when this technology meets actual human beings who have contracts and agents and Screen Actors Guild cards and thirty years of work that can now be voice-cloned in four minutes.

That gap — between "here is a thing that exists" and "here is what happens to real people when the thing exists" — is where the interesting work lives, and almost nobody in tech is doing it.

So we get our best material from Hollywood. Which is fine. I'll take it wherever it comes from.

The broader applicability isn't incidental. It's structural. The entertainment industry is just further along on the specific problem of "AI that produces output that competes directly with human labor in a domain where quality is subjective and attribution is everything." Every industry is going to get there. Entertainment got there first because the output is visible — you can watch the movie, you can hear the song, you can immediately tell whether a voice is real — and because the people affected were organized enough to fight back in a way that generated documentation.

The document exists. Someone made it. That someone was not us. This is the correct outcome — let the people with the most skin in the game do the primary research, and read carefully when they publish it.