There’s plenty to chuckle about when the occasional story on social media highlights what happens when a team of researchers feeds dozens of scripts from a television show to an algorithm and asks it to produce a new script based on its learnings. After all, it’s not like an AI-generated episode of Stranger Things could possibly get any more convoluted or unreal than the people-written genuine issue, though the results tend to be laughable and only mildly disconcerting.
But when it comes to actual movies and television shows, we may not be that far from AI setting the thematic guideposts and general structure for new scripts that people would want to watch. In fact, that’s the likelihood seen by Jaidev Amrite, SparkCognition’s senior director of product management, whose focus includes natural language processing (NLP).
“My wife and I have a common joke when we’re watching movies on Netflix. I will pause the movie and ask what’s going to happen next? And we have a running bet for how many times I’m right. So I would not be surprised if a lot of script writing, at least in terms of themes, is able to be done using AI,” he said.
“Putting in scripts of 20 episodes of Lost and generating a new script is somewhat amusing. But it wouldn’t take much work to codify some of the core ideas of back episodes of something like Mission Impossible, the TV show, and generate the broad narrative plot line for a new Mission Impossible movie. That actually is something that you could do.”
Movies and television shows are just one frontier in the world of content where AI has the potential to create massive change. While entertainment content generated with algorithms isn’t a huge ethical concern—except for the human creators who work in that field—there are a whole host of questions connected to informational content and news stories written by AI.
In fact, it’s a priority concern for search engine giant Google, with the company’s own research into AI being used to help it detect written material that may violate its guidelines against content created in large part by repurposing existing sources.
Amrite noted that, for years, AI has been useful for three main tasks related to content creation:
- Summarization: finding the most distinct pieces of information from a piece of text.
- Paraphrasing: restating a body of text in different words.
- Natural language generation: providing a seed topic or list of bullet points and asking AI to generate text around that core idea.
Those capabilities allow for the creation of content that doesn’t rely on human taste or nuance and can result in AI-generated news summaries, financial reports, basic recaps of sporting events, and other commoditized informational content. Beyond the world of news, industries such as law, health care, and financial services are using AI based on those basic abilities to dramatically reduce their reliance on human analysts.
The next realm for content involving AI, Amrite said, will come from recent advancements in computing power for cloud-based tools that have “leveled up” the ability to learn context and communication cues and will continue to improve the results for content created with the goal of matching real human interactions.
Those advances have interesting implications in the world of advertising and marketing: potentially helping brands identify the messages that are the most effective at influencing their customer’s buying decisions.
Continued Amrite: “Think about the notion of focus groups. If you took the output of all the focus groups that have ever been done, then you can build a pretty good data set. If you even look at just one agency and you looked at the outcome of all of the focus groups that they did, or if you looked at the outcome of all of the A/B testing that a web team might do with its copy, the AI can start ascribing qualities to certain words, in certain terms, to where copywriting could become more and more data-driven.”
Amrite said what’s likely to happen in the near future is that AI technology will continue to be trained with more powerful summarization and natural language processing. At that point the technology will be able to generate the core concepts of an initial content idea, or provide a starting point for a finished product that real people customize to make sure it connects in a genuine way.
“When it comes to human subjectivity and taste, there is very little in the way of AI that can address that question completely,” Amrite concluded. “As long as the consumer of something is human, there is going to be a premium on what humans consider to be important.
“AI is not automatically going to tell you which headline is the most meaningful for a person, and AI is not going to be able to say objectively which way of describing creates empathy, caring, and concern in a human who is receiving that information. So they’ll never probably be what you would call a complete solution to the problem of writing or communicating.”
And it’s not like AI could do any better than that hilarious Stranger Things joke at the top of this article. That, friends, is what we call job security for a writer!
