According to the thirteenth of Pixar's 22 rules of storytelling, characters must have opinions: "Passive/malleable might seem likeable to you as you write, but it's poison to the audience." But what if that character is a robot? That's what the author of the rules, Emma Coats, has been exploring as she works to give Google its new personality. Read more: Google Allo will now turn your selfies into Bitmoji-style stickers
Coats, 31, writes the dialogue for Google Assistant, the chatty digital helper the company is using to turn search queries into conversations. Unveiled in messaging app Allo in September 2016, then extended to the Pixel and Amazon Echo rival Home the following month, Assistant is intended to be the character at the core of Google's products - its AI-powered answer to Siri.
But a character requires a personality. That's what Coats, who joined Google in January 2016, was hired to create. "What I thought was really crazy and interesting is that you have to think of everything a user might ask," she says. "Be able to create this completely well-rounded character, be able to handle questions from any direction and come across as a consistent persona."
To construct Google Assistant's "easygoing, friendly" personality, Coats's small team in Mountain View, California - part of a division run by Google Doodle head Ryan Germick - imagines likely questions, then comes up with a range of responses, which are then handed over to the developers to code. The low-tech method harks back to Coats's time at Pixar, where she worked from 2006 to 2012 as a storyboard artist - mainly on Brave, but also on Monsters University and Inside Out. It was there that she wrote the 22 rules of storytelling, which started out as a series of widely shared tweets. Coats, who went freelance after leaving Pixar and before joining Google, describes the set of unofficial rules as "Notes to my younger self".
The nuggets of advice range from "Pull apart the stories you like" to "Story is testing, not refining" and "Come up with your ending before you figure out your middle." Read more: Google Home is coming to the UK on April 6 for £129
Not all of the rules apply to Google Assistant, however, for one simple reason: unlike in a film, this character isn't the hero. "You, the person interacting with it, are the hero," Coats says. That's why the Assistant can't be opinionated: it's there to be reliable, not to have depth. "If we gave it some dark conflict secret, that probably wouldn't be a great user experience." In Pixar terms, it's the "fun, trusty sidekick": Slinky Dog, not Buzz or Woody. Coats, however, has a different comparison – the pop star Demi Lovato.
After leaving Pixar to go freelance in 2012, Coats worked on a mobile game called Demi Lovato: Path To Fame. “Demi Lovato takes you, the main character, under her wing, and you rise to fame recording songs and making Demi proud,” Coats recalls. “One interesting thing that I learned from that, coming from a linear story background onto this interactive story, was that we did have an endearing sidekick. She never cuts you down, she's always open to your ideas, she's always excited about the things that you as the character choose to be excited about, and of course she always has really witty quips.”
Witty quips are one of the most crucial elements of Coats work at Google. Partly that's because humour is such an effective ways of building character. It's also because humour can be used to deflect awkward questions, especially for an AI that is learning its way. "We don't want to have to fall back on something like, 'I don't understand'," Coats explains. "That draws the attention back to you instead of continuing the conversation you're building." Ask the Assistant if it's human and it will say, "Well, I've been told I'm personable." Ask if it can learn and it will reply: "Learning is my jam," followed by a honeypot emoji.
All this is very cute, but it's not really the answer - because in fact the Assistant is learning: to displace its human writers. As it takes requests and listens to the user's reactions, its machine-learning algorithm improves itself, a process Google refers to, rather sinisterly, as "the transition".
By keeping users talking, Coats is putting herself (and perhaps, one day, all of us) out of work. Does this worry her? If it does, she doesn't show it. "I'm sure we'll be out of a job at some point as the Assistant learns faster and faster," she says, "but for now it's a really fun job."
This article was originally published by WIRED UK
