Artificial Intelligence and the Art of Storytelling
The human brain is wired for storytelling. As kids, we create make-believe worlds and instinctively tell stories. These stories don’t just entertain and educate us but also help in shaping our intellect and behavior. As adults, we continue to be immersed in stories that we read in books and magazines or watch in TV and movies. In fact, the foundation of creative arts is storytelling, be it through theatre, music or personal blogs.
Stories reflect the society and culture that they are produced in. A good story keeps you on the edge of your seat and evokes powerful emotions. It wins hearts, changes lives, and even has the potential to change the course of history. As technology advances, storytelling has also evolved. The next logical step in the evolution of storytelling is the introduction of artificial intelligence. While storytellers have long used artificially intelligent algorithms to improve their writing skills and enhance their stories, it’s a different ballgame when it comes to understanding stories and creating and developing storylines.
AI Content Creation
The plots for most stories follow the famous three-act structure which goes as follows:
- The protagonist is introduced and identifies the problem
- The protagonist attempts to resolve the problem
- The protagonist solves the problem
To put it simply, there is a plot, conflict, and resolution. The mix of these elements enables the dramatic arch of the story to evolve in a way that people can easily understand it. However, the question is that if a computer is introduced to this structure, would it be able to create relatable stories. Based on the cognitive analysis of inter-related elements, is it possible to create a coherent story that transfixes people? This is what researchers experimenting with a new-class of machine learning software and tools have been trying to decode.
How Artificial Intelligence Is Creating New Ways of Storytelling
There are many factors that have contributed to the success of AI in storytelling. The first is the development of the many-layered neural network and secondly, the creation and availability of huge databases that can power these networks. In 2016, the screenplay for the short film, Sunspring was written by Benjamin, a self-named system-on-chip running a long short-term memory recurrent neural network. By feeding a text recognition software multiple sci-fi movie scripts, the neural network could predict the words and phrases that most often appeared together and write the screenplay and the theme song for the film. While the film dialogues and script were touted as being “weird”, it did show a rapid progression of AI in storytelling.
According to Yves Bergquist, the CEO of AI research company Corto, “We are standing on the cusp of a massive revolution in how media companies will be able to tell stories. It’s about telling stories that really resonate with people and when the stories are right, media companies are going to take over the world. They are going to become much more powerful and successful.”
Researchers from UC Santa Barbara recently developed a neural network which can be used to deduce original, abstract stories from images. The AREL (Adversarial REward Learning) framework was not only able to come up with its own stories but also passed the Turing Test* three out of five times. While the sophistication of these machines is questionable, it could lay the groundwork for future networks.
AI Programming and Storytelling: Challenges
The major challenge for neural storytelling is the creation of databases for complex storytelling and the identification of the emotional arcs in stories. While most natural language processing software can crunch data and create word-by-word summaries, they’re severely lacking in the ability to generate several paragraphs of coherent text and map out a good story plot.
Prof. Patrick Winston, a renowned artificial intelligence expert at MIT, believes the “fairy dust” that separates humans from other animals is the ability to understand and tell stories.
One of the significant challenges for machines is that they have a hard time identifying emotional arcs that the characters go through such as situations fraught with danger and hardships, falling from grace, or declaring victory over evil.
However, researchers have been working on remedying this situation as well. To resolve the lack of emotional arcs in machine-generated storylines, a group of researchers from the University of Adelaide and the University of Vermont collected computer-generated emotional arcs for nearly 2000 works of fiction. Based on the outcome of the protagonist in the situation, they then classified these into six core types of narratives, which includes Rags to Riches, Cinderella, Icarus, Riches to Rags, etc.
Instead of focusing on the plot, the researchers used sentiment analysis to generate the emotional trajectory of the story along with the emotional structure that the story writers were most likely to use. This was then contrasted with the structures that the readers liked the best.
Can AI Replace Human Storytellers
While the scenario may seem grim for content creators and storytellers, experts suggest that there is still a long way to go before robots can tap into the secret sauce of every successful story – empathy. Even after knowing the storytelling archetypes, there is no tried and tested storytelling recipe that would evoke emotions and attract viewers. It’s impossible for AI to create memorable, emotional stories to engage their audience – at least for now.
Instead of treating it as a threat, AI can be used to create better content and eliminate many of the everyday tasks. It can serve as a valuable tool for writers and storyboarders to amplify the emotional pull and sharpen stories.
*Turing test was developed by Alan Turing in 1950 to test a machine’s ability to exhibit intelligent behavior, equivalent to that of a human.