Since it began to sound strongly in the media, ChatGPT has become one of the fashionable topics and a tool —that’s how they want to sell us— called to revolutionize productivity of human beings or, according to the most catastrophic, to destroy jobs and end up replacing us in the not too distant future. But, as the meme says, “let’s calm down”.
The other day, listening to the episode of the ‘Scriptnotes’ podcast entitled ‘Rian Johnson Returns’, I could see between laughs how this artificial intelligence is much less intelligent than it might seem at first glance. In it, John August and Craig Mazin took advantage of the return of the director of ‘The Last Jedi’ to test the AI with a most peculiar request.
Daggers in the Back: Fudge in Space
At one point in the program, the trio of writers had the idea of ask ChatGPT to create a premise for a third party of ‘Puñales por el espalda’ set in space using the following prompt: “Give me a premise for a ‘Knives in the Back’ murder mystery set in space with Benoit Blanc”. Be careful, because the result is not wasted.
“In this futuristic murder mystery set in space, renowned detective Benoit Blanc is called in to investigate the apparent suicide of a wealthy space station owner. Arriving at the space station, Blanc quickly realizes that the death was in A carefully planned murder comes to life and sets out to uncover the truth. With the help of a diverse cast of suspects, including the victim’s family, Blanc must untangle the complex web of lies and motives to identify the killer. As he delves deeper into In any case, he finds himself up against a cunning adversary who will do anything to keep his deadly secret hidden.
The first reaction that we can have when reading the proposal diagonally, like Mazin, August and Johnson, is to be surprised at what, at first glance, is a perfectly constructed text. But as soon as we do a more detailed reading, we can realize that the text is nothing work of what the showrunner of ‘The Last of Us’ aptly describes as a bullshit artist —a faker, come on.
“I love how what this thing is doing is basically a great job being a fake. I’ve seen this in writing rooms where people start throwing ideas around and it’s all based on cadence and trust. Then you realize account: ‘They have not told me anything.’ From my point of view ChatGPT does not seem to me that it is a threat to the writers. It seems to me a threat to the producers who go around saying things like this”.
Later, Mazin stated that the basic proposal of ChatGPT reminds him “to the understanding of the narrative structure that I had when I was starting”but then they decided tighten the screws on the AI a little more with a new prompt: “Okay, this doesn’t seem too space-oriented. Let’s make sure it’s very space-focused, with things that can only happen there. Give me a three-act structure.”.
The result, to no one’s surprise, was equally botched, with the AI limiting himself to reformulating his first result, coping with the need to expand the text by arming himself with equally empty clichés. For example, this button in which he synthesizes the third act, which becomes almost an episode of ‘Battlestar Galactica’: “In a gripping finale, Blanc must pursue the killer across the vastness of space to bring him to justice”.
At this point in the experiment, it was clear to me and its makers that the writers’ job is not in jeopardy, but still, Rian Johnson believes that something like this could work in a real situation.
“You can imagine this working perfectly in a writers room. Some jerk stands up at the desk waving his arm and pointing at the executive [mientras hace el pitch]. He starts to pick up the water bottle from the table and throws it. You can see executives walking out of the room saying, ‘That’s the best pitch I’ve ever seen’.”
And, then, with the logical doubts about where this technology will go over time, Craig Mazin labeled the current ChatGPT within the world of film writing. And no, it’s not a productivity-boosting tool, a secret weapon to make our lives easier, or some sort of Skynet that will end the world of work as we know it: for good Craig, “ChatGPT is an asshole producer”.
The dialogue scene
If you still have doubts about what has been exposed so far, hold on, because curves are coming. And it is that after the premise and the structure in three acts, it was time to ask the AI to develop a dialogue scene in which Blanc met Constance Pickering; a rival researcher… with whom he soon allys himself with not even four lines of —embarrassing— dialogue.
“Blanc is standing in front of a holographic display from the victim’s spacesuit, examining a ruptured air hose. Pickering walks over and clears his throat, drawing Blanc’s attention. ‘I’m Constance Pickering, and I’ve been assigned to investigate the death. of Marcus Wells.’ Blanc raises an eyebrow. ‘Marcus Wells? I was under the impression his death was a suicide.’ That doesn’t sound like something I’d say. Pickering shakes his head. ‘Not according to the evidence. broken air in his spacesuit indicates that he was suffocated’.
‘This was clearly a murder.’ Blanc considers this for a moment. ‘Interesting, and what clues do we have so far?’ Pickering crosses his arms. ‘So far, we have a list of suspects but nothing concrete. I was hoping to discuss the case with you and see if we could combine our efforts.’ White nods. ‘I think that’s a splendid idea.'”
The initial reaction of the trio of filmmakers was obviously to get mad and throw their hands up. Nevertheless, John August made a very interesting point about the abilities of ChatGPT with which his colleagues agreed:
“What is surprising and impressive is the fact that this language model understands that a murder in space may be related to a broken air hose and that suffocation may be one of the reasons.”
And now that?
After going around the obvious, it is time to put an end to this peculiar exploration of the “specific abilities” of our friend GPT, nothing better than closing August’s conclusion about why this artificial intelligence has nothing to do against human beings. A weakness that, precisely, has a lot to do with its nature.
“Screenwriting is about putting words on the page, which ChatGPT is getting better at, but it’s also about being able to be with people and communicating what your vision is and being able to respond and address their concerns. It’s about of being able to talk to Daniel Craig about why his line is that way and how to work it, if it’s not a good fit for that actor or the moment.
Much of the work is not just the words on the page, but also the communication about the story as a whole. That’s what computers have a much harder time doing. It’s the subtle thing, having to be there in person, you have to be there in the flesh to do things. That will not change”.
Creatives of the world, peace of mind: we are still essential.