Google’s AI Predicts the Next Sentence of Dead Authors

artificial intelligence google shakespeare ray kurzweil
A portrait of William Shakespeare engraved by Martin Droeshout in 1623. Writings from the famous playwright were used by Google to train an artificial intelligence algorithm to predict writing behaviors. Wikimedia Commons

Google has developed an artificial intelligence system capable of determining the personality traits of dead authors and predicting what they would write next.

The Natural Language Understanding research group, led by artificial intelligence pioneer Ray Kurzweil, built software able to understand natural language beyond just the definitions of words and rules of grammar.

The system developed aims to give computers context, as well as an understanding of the writer’s personality and language patterns, in order to interpret language in a similar way to humans.

“One way to explore this research is to build a system capable of sentence prediction,” Google researchers Marc Pickett, Chris Tar and Brian Strope, wrote in a post on Google’s blog. “Can we build a system that can, given a sentence from a book and knowledge of the author’s style and ‘personality,’ predict what the author is most likely to write next?”

The research drew upon the work of 1,000 different writers, including Charles Dickens, James Joyce and William Shakespeare. Sentences from the writers were fed into a Deep Neural Network in order to train it to predict what sentence might come next based on what has previously been written.

Researchers were also able to use the method to test writers using the Myers Briggs personality test, determining Shakespeare was more of “a private person,” while Mark Twain was more of “an outgoing person.”

If developed further, the system could be used to better understand the personality of Shakespeare, whose private life remains the subject of academic speculation due to very few records remaining. Ultimately, the technology could hold the potential to offer possible conclusions to unfinished novels from writers like Albert Camus, Ernest Hemingway and David Foster Wallace.

“Combined with our generative model, these vectors allow us to generate responses as different authors,” the researchers wrote. “In effect, one can chat with a statistical representation of the text written by Shakespeare!”

Other applications for the technology include the generation of automated email responses based on an individual’s writing style. Google has already used machine intelligence to provide personalized replies through Gmail’s Smart Reply feature.

“This work is an early step towards better understanding intent, and how long-term context influences interpretation of text,” the researchers concluded. “In addition to being fun and interesting, this work has the potential to enrich products through personalization.”