Arts – Novel approach
Will computers one day have the emotional intelligence and sophistication to write fiction? Ry Crozier investigates.
Imagine switching on a computer and having it independently churn out a bestselling novel replete with plot twists and complex characters, showcasing the full suite of human emotions.
While such an intelligent author-computer is still a long way off, Australian Research Fellow and artificial intelligence expert Dr Malcolm Ryan predicts “computers will be making interesting and meaningful contributions to literature within the decade”.
“They might be more experimental than mainstream,” concedes Ryan, who is based in the School of Computer Science and Engineering, “but the computer will definitely be doing some of the work of writing.”
While today’s computers are very smart, they still can’t compete with the human brain in terms of imagination and creativity. “Most people can tell a story without having to consciously think about what goes into it – the characters, events, sequencing, language and level of detail,” says Margaret Sarlej, one of Ryan’s PhD candidates.
“That’s an innate talent that computers just don’t have. They need detailed instructions for every step of the process, which is not easy to provide when, to a large extent, we don’t even understand how people do it,” she says. It’s easy to underestimate just how much effort is required to make computers more human-like and autonomous, Ryan says.
The enormity of the challenge was apparent in 2007 when he tried to get a computer to understand, and then reproduce, a page from Beatrix Potter’s children’s classic The Tale of Peter Rabbit. Though the story-line appeared straightforward, Ryan found the level of complexity in the characters and their emotions was simply beyond what the artificial intelligence at the time could handle.
“Humans have remarkable ability to do complicated things and see them as being quite simple,” he says. “It’s only when you have to program a computer to do the same thing that you realise how much knowledge is required.”
Rather than show off the computer’s limited storytelling abilities, Ryan set about describing everything it couldn’t do. “It was a far more instructive approach,” he says.
Seven years on, Ryan and his PhD students believe they are at the stage where they can concentrate their research on what can be done.
Sarlej recently developed a program that allowed a computer to write simple Aesop-like fables. The program – the MOral Storytelling System (MOSS) – resulted in stories more “alive” than previously considered possible. It structured a plot around characters imbued with up to 22 different emotions, allowing multiple plot development possibilities and the creation of coherent, interesting and useful stories in their own right.
Ryan and Sarlej see it as a significant advance in capability. “While it’s still very early stages, it’s an important step up for what these artificially intelligent storytelling systems can do,” Ryan says.
The group is hoping authors, game designers and other creators will see the value of the research and get on board.
“For us this is a serious literary project, and we want to find artists who can help direct it to that end,” Ryan says. “How will this technology be used? It is impossible to predict. We hope artists will take it up and create things we’d never even imagined.”
What the computer wrote about retribution …
“Once upon a time there lived a unicorn, a knight and a fairy. The unicorn loved the knight. One summer’s morning the fairy stole the sword from the knight. As a result, the knight didn’t have the sword. The knight felt distress that he didn’t have the sword anymore.
“The knight felt anger towards the fairy about stealing the sword because he didn’t have the sword anymore. The unicorn and the knight started to hate the fairy.
“The next day the unicorn kidnapped the fairy. As a result, the fairy was not free. The fairy felt distress that she was not free.”