Machines may be smart, but they make pretty dull companions. Google knows this, and as it builds out its recently announced Google Assistant personal assistant technology—which, like Apple's Siri and Amazon's Alexa, will answer questions and follow commands like "turn on the lights"—the company is eager to make artificial intelligence more personable. How? By throwing some artists at the problem, naturally.
Ryan Germick, who heads the Google Doodle team in Mountain View, is now working on adding a human touch to Google Assistant, along with Emma Coats, a freelance artist best known for her stint as a story artist at Pixar and the author of a viral list of 22 rules of storytelling. In a session at the Moogfest music and technology festival in Durham, North Carolina, this weekend, Germick and Coats talked about the process of making A.I. feel friendlier.
Google Assistant is a new take on the artificially intelligent "personal assistant" concept that will be built into Google products like the forthcoming Google Home speaker. Like Siri and Alexa, Google Assistant will take the form of a conversational interface, rather than any sort of humanoid-looking robot—at least for now. To make these human-to-machine exchanges more enjoyable, Google is tapping creatives like Germick and Coats to inject some humor and storytelling smarts into the conversations.
"Because it’s out in the wild and people can say anything to it, we have to create the most well-rounded character that we ever have tried," Coats says. "One of the things that we’re working on is how to make it relatable. How does the character think of itself in a way that you can relate to? What is its childhood?"
It's no slip of the tongue that Coats refers to Google Assistant as "a character." Her experience working on Pixar films like Brave lends itself perfectly to the less-robotic type of A.I. that Google is hoping to build into its products in the future.
To make Google's machines seem more relatable, Google is looking at building backstory narratives into Assistant's personality so that people might be able to draw comparisons between their own lives and these artificial beings. They're also experimenting with things like adding an element of surprise into the bot-human exchanges and building trust by making the A.I. character seem vulnerable—like the bot needs our help from time to time.
The next generation of Google's A.I. hopes to take on a more casual, companion-like vibe than we're used to with Siri and Alexa. Part of that means programming with answers that are purely fun. For example, asking Google Assistant its favorite flavor of ice cream could yield an answer like "You can't go wrong with Neapolitan." Asking the machine "Do you fart?" might lead it to say "Not recently," or perhaps something more sassy. In time, Germick jokes, Google could ping data about nearby air quality test results.
We spend a lot of time with our phones and the connected gadgets that increasingly fill our home, such as Amazon's Echo (or, Google hopes, its upcoming Echo competitor, Google Home). For Coats, Germick, and the other people working on Google’s new A.I., it’s all about making their underlying artificial personality feel like somebody you’d want to hang out with all the time. Because, well, you could be.
"You become friends with people because you spend a lot of time with them," Coats says. "There have been studies that show that it doesn’t necessarily matter if you instantly click with someone. One thing we do is to make the character as entertaining as we possibly can, so that you want to spend time with it."
It's still quite early for Google Assistant, which was just announced at Google I/O last week, so concrete details are scarce. But suffice it to say that when Google unleashes Assistant into the wild, they'll have done their best to make it feel less like a soulless robot and more like an old friend.