No it's not. You're conflating some highly different concepts.
Generative AI will not for a long time write screenplays, although Causality is able to do a lot better at writing scenes because we know a lot more about the story, so we can write a first garbage scene that'll get you past the blank page, and occasionally provide good ideas. Given the right context provided, we've had it write scenes that we had to agree were not totally bad. But it's only because you've already spent a lot of time explaining the story in terms of tags, synopsis etc, and it just fleshes out half a page of what is your story. Still, no one is imagining that a single word of what is generated will make it to any final screenplay. Subtlety is utterly lost on AI.
But the most natural fit for large language models is natural language processing, and Causality already does a spectacular job at writing beat titles and synopsis based on what you've written. I've basically stopped writing summaries myself, because the summaries written by Causality are genuinely good.
The change location feature also works remarkably well, and is again a natural fit. It'll change a scene in a car to a scene in a diner, making mostly the right changes.
The novel to screenplay and screenplay to novel also works remarkably well. We have a person who's adapting a novel to a screenplay, and it turns out that most of the work is a language conversion the AI does very well. You convert a beat, and then do an immediate human rewrite. It just shovels some of the sand for you, and it does it well.
Where it breaks down is around emotions, because everything the AI does is cartoony. If you ask it to make a character gradually more angry during a scene, it'll just start adding "(angry)" parentheticals. And in general large language models do a poor job at rewriting whole text, because it's bad a preserving the existing text. We're finding greater success processing each paragraph separately, where we're able to do more subtle prompts, because we don't have to preserve the scene around it. Then we start to be able to increase or decrease a certain emotion, and the result isn't bad. But it still needs immediate human editing, and emotional changes are the lowest on the list, and not a good job for AI.
We're mostly interested in AI to do the brunt work of language processing.
And then also having the AI write garbage drafts of scenes based on your descriptions and place in the story arcs, is extremely helpful. Writer's block is about wanting the first attempt to be perfect, but with a bad draft already on the page and burning your eyeballs, you're motivated and confident you can do better. You can also use AI drafts like a Tarot deck, to give you oddball ideas.
So don't be so quick to dismiss. It doesn't appreciate many important differences. AI isn't one thing. We can all agree that generative AI writing is crap, and will be for a very long time. But natural language processing is real, and AI does a far better job at it than the previous algorithmic approaches of the last 50 years, well enough that you can have a personal writer's assistant to do some brunt work that's not interesting or creative.
No it's not. You're conflating some highly different concepts.
Generative AI will not for a long time write screenplays, although Causality is able to do a lot better at writing scenes because we know a lot more about the story, so we can write a first garbage scene that'll get you past the blank page, and occasionally provide good ideas. Given the right context provided, we've had it write scenes that we had to agree were not totally bad. But it's only because you've already spent a lot of time explaining the story in terms of tags, synopsis etc, and it just fleshes out half a page of what is your story. Still, no one is imagining that a single word of what is generated will make it to any final screenplay. Subtlety is utterly lost on AI.
But the most natural fit for large language models is natural language processing, and Causality already does a spectacular job at writing beat titles and synopsis based on what you've written. I've basically stopped writing summaries myself, because the summaries written by Causality are genuinely good.
The change location feature also works remarkably well, and is again a natural fit. It'll change a scene in a car to a scene in a diner, making mostly the right changes.
The novel to screenplay and screenplay to novel also works remarkably well. We have a person who's adapting a novel to a screenplay, and it turns out that most of the work is a language conversion the AI does very well. You convert a beat, and then do an immediate human rewrite. It just shovels some of the sand for you, and it does it well.
Where it breaks down is around emotions, because everything the AI does is cartoony. If you ask it to make a character gradually more angry during a scene, it'll just start adding "(angry)" parentheticals. And in general large language models do a poor job at rewriting whole text, because it's bad a preserving the existing text. We're finding greater success processing each paragraph separately, where we're able to do more subtle prompts, because we don't have to preserve the scene around it. Then we start to be able to increase or decrease a certain emotion, and the result isn't bad. But it still needs immediate human editing, and emotional changes are the lowest on the list, and not a good job for AI.
We're mostly interested in AI to do the brunt work of language processing.
And then also having the AI write garbage drafts of scenes based on your descriptions and place in the story arcs, is extremely helpful. Writer's block is about wanting the first attempt to be perfect, but with a bad draft already on the page and burning your eyeballs, you're motivated and confident you can do better. You can also use AI drafts like a Tarot deck, to give you oddball ideas.
So don't be so quick to dismiss. It doesn't appreciate many important differences. AI isn't one thing. We can all agree that generative AI writing is crap, and will be for a very long time. But natural language processing is real, and AI does a far better job at it than the previous algorithmic approaches of the last 50 years, well enough that you can have a personal writer's assistant to do some brunt work that's not interesting or creative.