Can AI write an episode of Stargate? Google AI rose to the challenge

The process of writing a television show usually involves a writer’s room and a lot of time as people come up with the plot and dialogue that make a show work.

For the cult classic Stargate science fiction franchise, which spanned three series (SG-1, Stargate Atlantis, and Stargate Universe), character and plot development was overseen by Stargate co-creator Brad Wright. In 2021, Wright publicly posted a message on Twitter asking if it was possible for AI to write an episode of Stargate that would appear on the SciFi insider site the companion†

None other than Laurence Moroney, AI leader at Google, responded by taking up the gauntlet of trying to prove what AI could do. Although initially he was not afraid that AI would replace him or other writers.

“When the whole project started, I pitched it to The Companion as an idea — I knew I’d seen a few AI models for scripts,” Wright told VentureBeat. “In some ways it’s fantastic. In other respects, it’s very harmless.”

The first iteration of the AI-generated script was completed in November 2021. The script was interesting, but there was also a lot of gibberish, Wright said. Now Moroney and Wright are teaming up again on a second draft that aims to dial a new port address for a more involved and engaging Stargate script.

“I told Laurence (Moroney) that if we’re going to do this again, we’ll probably have to try and really step up the game, and he accepted that challenge,” Wright said. “That’s what blew me away, because it’s not just better – it’s like, ho, better!”

How the Stargate AI script was generated

The process for producing the Stargate AI-generated script was much the same as the way any AI model is first built – by training it.

Moroney trained the AI ​​model with every Stargate episode script ever written, giving the system a corpus of every line of dialogue and plot. He used several technologies, mainly Google’s TensorFlow machine learning framework.

He also used pre-trained natural language models, along with a technique known as transformers, Moroney said.

“Not to be confused with the Hasbro toy, the technique called transformers was invented primarily for language translation,” Moroney said. “When you think about the idea of ​​language translation, you have an input sentence that you want to match with an output sentence and that almost sounds like the ideal thing for script generation.”

For example, an input phrase might be “Captain Samantha Carter says something to General Jack O’Neil.” The output phrase is derived from the Transformer’s training on how Jack has responded in the past to input phrases similar to the one Samantha just uttered.

“So I could train a transformer to predict what Jack would say to something,” Moroney said.

The other core technology used was something known as a universal sentence encoder, which gives a numerical value for the context of a sentence. With that approach, Moroney said it was possible to encode semantics in sentences numerically, to understand the connections between one sentence and the next, moving forward and backward. Wright noted that, in his opinion, the second version of the Stargate API was better than the first because of the encoder – since the script never fell into nonsense gibberish.

While there are several Google machine learning tools that Moroney could have used for more automation, he stressed that much of the process of building the Stargate script was manual. He gave a prompt to the trained TensorFlow model, which would then respond. Those responses were then fed into the script, which Moroney put together manually.

“The model didn’t spit out a scenario with properly formatted text,” Moroney said. “It was a couple of models that generated the right tags to get in at the right time.

Bringing scripting AI to the Google enterprise

The models and techniques Moroney uses to develop the Stargate script may be applicable to the enterprise. † However, he joked that his AI techniques are unlikely to be used to write an event keynote script for Google and Alphabet CEO Sundar Pichai anytime soon.

“If I could hear every word Sundar (Pichai) has ever said, I could train a model in his vernacular,” he said. “If I were to write something that could help my pitch so that it’s more in his voice than in my voice, things like that would be a very helpful resource.”

Moroney said that for writers who help produce the scripts used for keynotes at a corporate event, an AI model can help reduce the time it takes to get the right tone for a particular speaker, but it won’t. can do all the work.

“When you make an announcement for something like Google I/O, you’re talking about something new that no one has seen before,” he said. “So by definition you don’t have the data to train such models.”

Another possible way AI is now being misused is to produce deep counterfeits, which pose a non-trivial cybersecurity risk. An AI trained in the correct vernacular could potentially be used to aid in the production of a deep fake, but Moroney said he prefers positive uses of technology over negative ones.

“Personally, I’m very concerned about deep fakes,” he said. “But I’m also optimistic that there are technologies that you can use to detect deep counterfeits and they are actually quite easy to spot with the right application of technologies.”

Stargate AI script is good, but it won’t replace humans

A common concern about AI is that it will replace humans, but Moroney isn’t worried for now. According to him, AI-assisted tools could potentially help lay the groundwork for future scripts. While humans can use AI to help with parts of the process, the core stories and uniqueness come from things that cannot be trained by machine learning.

Wright emphatically said that the AI-generated script is a unique idea that honors the Stargate show, but it honors the show as it was. That said, as a man who has spent his entire career writing about the future, there is always the potential for something more.

“I think eventually AI will be able to generate something that will pass a test of whether an audience believes it was written by humans,” Wright said. “It may not be better than the best people, but you know, in the beginning the best chess masters beat the best computers and now AI beats the chess masters. I’m sure something similar could happen with this.”

The Stargate AI version 2 script, as read by the original Stargate actors – including Richard Dean Anderson, Amanda Tapping and Michael Shanks – will premiere on the companion on May 21.