May 8, 2016 weblog
Google feeds romance fiction into AI engine
When was the Eiffel Tower built? Whether it is your brother or a screen giving you the date, it makes no difference. You need to know the date, regardless.
How do you make the results of the work of neural networks, though, sound as if a person was speaking to you and not a machine if what you express wants a different type of answer? Getting that machine to relay a response that approximates real human conversation with "personality" is a challenge for those engaged in AI research. Nonetheless, the researchers are not about to give up any time soon.
Google made news recently with reports on how it wants to teach neural networks more conversational language skills using romance novels.
Popular Science reported how Google has consumed over 2000 of the novels in order to create more conversational exchanges with people.
Romance novels is the genre that was selected because the language is so conversational. The Google team in charge of all this has been feeding this type of fiction into the engine "to expand its vocabulary and creative abilities. The AI system then writes sentences and compares them to those it has read," wrote Lindsey Smith in The Verge.
San Francisco-based Alex Kantrowitz, a senior technology reporter for BuzzFeed News, provided more detail into just how this works: "The neural network starts off without any knowledge base, so feeding it text from a book is like reading a novel to a baby and hoping it picks some of it up. This is why it requires reams of data—or about 2,865 romance novels—on which to build its so-called intelligence. After ingesting those romance novels, Google's AI engine composed sentences of its own using what it learned from them. It then evaluated these new sentences against the original text. The process was repeated over and over again, with the AI self-calibrating as it went along—writing better and better sentences."
Oriol Vinyals and Andrew Dai are two key drivers behind the Google project. Smith said Andrew Dai wrote in an email to The Verge. "It's like how you'd rather ask a friend about what do to in a vacation spot instead of calling their visitor center."
Smith said, "Dai and his team simply hope that the AI engine will become more conversational, thus providing a better user experience."
Dai told BuzzFeed News: "Hopefully with this work, and future work, it can be more conversational, or can have a more varied tone, or style, or register."
Kantrowitz explained why romance novels was chosen and not something simple and conversational such as children's books. "Romance novels work better than children's learn-to-read books, since they offer a broad range of linguistic examples for the AI to draw from," he said.
© 2016 Tech Xplore