MOUNTAIN VIEW, Calif.— Google made its usual series of interesting and semi-interesting product and update announcements onstage May 8 on Day One of its I/O conference at Shoreline Amphitheater.
But it truly stunned the crowd of about 10,000 developers with these two: 1) a demonstration of Google Assistant conducting very human-like conversations in making typical haircut and restaurant reservations; and 2) the advancements Google has made in packaging and delivering news content.
We’re only a short time away from allowing our autonomous cars to make traffic decisions in transporting us around town; now it looks like we’re on the edge of letting digital assistants make dinner reservations for us—also without human intervention.
Google Assistant, though still in development, is becoming capable of making phone calls and having intelligent conversations for users the way an avatar might do it, only there’s nothing robotic or unnatural about how it works.
Google Assistant Plus Duplex AI is Impressive
CEO Sundar Pichai demonstrated two phone call recordings placed by Google Assistant—the first to a hair salon, the second to a restaurant. The voices—the first a woman, the second a man—sounded astoundingly natural; the humans on the other end hadn’t a clue they were talking to an artificial intelligence-powered digital helper.
The Assistant, in effect, was able to think for itself; in fact, in this instance it thought more effectively than the human in the restaurant reservation scenario. The Assistant was very clear, and in a nice way, about what day and time it wanted to reserve a table for a group of four people. The clerk at the restaurant kept forgetting the day, the number of people in the party and whether there might be a long wait for the group at that particular time. It was a very humanlike interaction all the way around, and the demo earned laughter as well as respect for the technology.
Pichai said that “this is the result of many years of work with natural language processing, machine learning, deep neural networks and other technologies.”
Google will conduct early testing of Duplex—the AI engine inside Assistant—this summer specifically to help users make the type of hair and restaurant reservations noted above. Once those use cases are locked down satisfactorily, the team will move on to other tasks.
Pichai said Assistant and Duplex can handle interactions gracefully and react intelligently, “even when a conversation turns into a misunderstanding or results in something unexpected.” In its current test mode, Google said the feature is able to complete most conversations successfully on its own without intervention from a person on Google’s end.
As one might imagine, there are still instances where it becomes confused and must hand off the task to a human.
Go here to read a blog post on how Google plans to continue developing Duplex with Assistant.
How Google News is Evolving
“We recently started our Google News Initiative and committed $300 million over the next three years to work with organizations and journalists to develop innovative and programs that help the industry,” Pichai said. “There is more great journalism being produced today than ever before. It’s also true that people turn to Google in times of need, and we have a responsibility to provide that information.”
So Google now is using AI to “bring forward the best of what journalism has to offer,” Pichai said. “We want to give readers quality sources they can trust, but we also want to build a product that works for publishers. Above all, we want to make sure we’re giving them deeper insight and a fuller perspective about any topic they’re interested in.”
When users now go to Google News, they now will see the day’s top five news stories–using the company’s vast knowledge about you—that Google thinks you would want to see. And Google knows a lot about its users, so it’s probably going to be pretty accurate most of the time.
“Google constantly threads through millions of articles, video, podcasts and web sites published every minute and assembles the key things you need to know,” Google Distinguished Engineer Tristan Upstill (pictured) said during the keynote.
Google News also pulls in local news and event in the user’s immediate area. You don’t have to tell the app anything about what you like to read, what your politics are, or what sports team you follow; Google AI already knows, and this all works automatically.
Of course that can be creepy, but we should all be quite used to this tradeoff of personal information for services by now. If you’re not, then stay with it or get out: the choice is yours.
It Can Be Invasive, But You Can Control It
At any point, users can tell Google News to show less or more of any publisher or topic. There are always controls.
When a user wants to see what stories the world is reading, he or she can switch over to Google Headlines to see general news, sports and features from around the world. The app also brings in YouTube videos along with everything else in real time.
Newscasts is another new feature that offers a preview of a story as it appears in a newspaper column, so that users can scan it more thoroughly than a headline and sentence or two below it.
Google’s Full Coverage view, using a technology called Temporal Colocality, is a literal wide-angle look at the day’s news, sort of like physically opening a newspaper to its full width. It collates the people, places, things and action in a story and presents it all in real time.
“This is by far the most powerful feature of the app,” Upstill said, “and provides a whole new way to dig into the news.”
Full Coverage presents a topic or single story from many different points of view and media types; users will see standard news stories, videos, podcasts, Twitter tweets, static images and others all on one large layout that’s really made for a laptop or desktop computer rather than a phone.
For example, a Full Coverage look at the recent power outage in Puerto Rico displayed stories about how it started, how could it have been prevented, if things are actually getting better, and so on.
