Google’s AI has been set loose on 11,000 unpublished books, including 3,000 steamy romance titles, in a bid to produce more naturalistic sounding vocabulary. As a result, the machine produced some surprisingly abstract and melancholy poetry.
In an unpublished paper entitled “Generating Sentences from a Continuous Space,” researchers documented what the Google Brain Team’s AI had learned from its romantic literature.
The team gave the AI a starting sentence and an ending sentence.
Then they asked artificial intelligence to bridge the two concepts using up to thirteen additional sentences.
The team gave it gave it a beginning and an end and asked it to tell a story.
In response, the AI has penned its own poems:
there is no one else in the world.
there is no one else in sight.
they were the only ones who mattered.
they were the only ones left.
he had to be with me.
she had to be with him.
i had to do this.
i wanted to kill him.
i started to cry.
i turned to him.
And another:
i went to the store to buy some groceries.
i store to buy some groceries.
i were to buy any groceries.
horses are to buy any groceries.
horses are to buy any animal.
horses the favorite any animal.
horses the favorite favorite animal.
horses are my favorite animal.
In the example above, the researchers gave the artificial intelligence a sentence about buying groceries and said fill in the gaps.
In the research paper the team explained the AI system was able to “generate coherent and diverse sentences through purely continuous sampling”.
Using an autoencoder, a type of AI network that uses a data set to reproduce a result (in this case writing sentences) using fewer steps, the team produced the sentences.
The principle behind the research is to create an AI that is able to communicate via “natural language sentences”.
This work could eventually lead to the development of a system that is capable of communicating in a more human-like way. Such a breakthrough is key in the creation of more useful and responsive chat bots and AI-powered personal assistants such as Siri and Google Now.
In a similar project, Google researchers have been teaching AI how to understand language by predicting and replicating the work of dead authors archived on Project Gutenberg.