Home » Science & Technology » Google is learning the language of sensitivity. From 2,865 romance novels

Google is learning the language of sensitivity. From 2,865 romance novels

Catch Team | Updated on: 10 February 2017, 1:50 IST

Take me, dammit, and free me.

But he didn't pull me close. Didn't press his hands to my ass and grind himself against my thighs. Didn't slam me against the wall and press his mouth to mine while one hand closed tight around my breast and the other yanked up my skirt.

He did nothing but look at me - and in looking made me feel as though he'd done all those things.

Smut on your mind, and your bookshelf? Quit guilt-tripping. Google's doing it as well - to improve its artificial intelligence (AI) engine, no less.

According to latest reports, the past few months have seen Google inject a dash of love and erotica from close to 2,865 romance novels (including


, from where the above extract is taken) into its AI. Other titles include the equally unsubtle

Unconditional Love, Fatal Desire

, and

Jacked Up


The team behind this project to 'sensitise' Google says the objective was to enhance existing technology, which is a bit staid.

The Google software engineer who led the project, Andrew Dai, along with Oriol Vinyals, told

BuzzFeed News

that, "In the Google app, the responses are very factual. Hopefully with this work, and future work, it can be more conversational, or can have a more varied tone, or style, or register."

"It would be much more satisfying to ask Google questions if it really understood the nuances of what you were asking for, and could reply in a more natural and familiar way," Dai explained to

The Verge

. "It's like how you'd rather ask a friend about what to do in a vacation spot instead of calling their visitor centre."

Any AI system is basically like a neural network that can adapt and evolve itself based on external inputs. The challenge becomes tougher when the inputs are specifically linguistic in nature.

Because a whole host of Google products depend on how computers respond to humans, feeding the AI more language seemed logical. But sensitivity, good grammar and vocabulary from steamy love sagas?

The real reason is actually more interesting, and it's something you've probably noticed often: the average romance novel basically has the same plot. "Girl falls in love with boy, boy falls in love with a different girl. Romance tragedy" says Dai. Same plot, different sentences and words? This makes this a perfect template for AI to assimilate. By constantly processing the 'same' story, the AI starts to detect similar phrases or meanings, helping increase its intuition levels.

Which is precisely what Google did to its AI - after getting 'fed' all those stories, it started developing its own sentences. Not just that, it also evaluated those sentences against the original ones from the novels to measure precision. The whole process was repeated till it became more and more finetuned and produced better - original - sentences, on its own.

Apart from the Google app, the improved AI technology could also be used for Google Inbox's "Smart Reply" product.

Dai has gone to the extent of saying that the AI could "theoretically" write a small book, given how many it's read.

What about debacles like Microsoft's genocidal chatbot Tay? "We work directly with the product folk on how to develop this with minimal risk of it doing bad things, things that we don't expect," says Dai.


Alex Kantrowitz asks a very pertinent question: sSince Google's AI has absorbed so much mush, is it possible to make someone fall in love with it? "It could happen eventually," accepts Dai. "There's an ancient Greek story about a guy who builds a statue of the most beautiful woman. The statue is more beautiful than any other woman, and he falls in love with the statue. If you can fall in love with a statue, I don't see why you couldn't fall in love with a neural network trained on romance novels."

The future looks set to be many things, but simple certainly isn't it.

Edited by Payal Puri

First published: 8 May 2016, 12:29 IST