Google Shows Off Its Latest “Baby Steps” Toward Becoming the Star Trek Computer
On Wednesday, Google announced new natural-language voice search on iOS, new, touch-friendly Web interfaces for its answers, and an ambitious, voluntary experiment to bring your Gmail messages into search results. Google wants to be like the Star Trek computer when it grows up, letting us ask questions naturally about anything and get a compelling answer right back. These new features show that Google’s much further along than you might realize.
English-Language Knowledge Graph Everywhere
Starting Thursday, users searching Google in English anywhere in the world will get Knowledge Graph results. Knowledge Graph, which launched in May, is Google’s effort to move away from simple keyword matching and toward a natural-language understanding of terms and concepts. This week U.S. users will also start seeing Knowledge Graph people, place and thing suggestions in auto-suggest and auto-complete while typing their queries.
Knowledge Graph boxes are displayed on the right side of search results and have a two-fold purpose: helping the you (and Google) quickly clarify the query by suggesting related objects and concepts, and helping you explore the relationships between concepts.
Not only does this visual representation of the things in your query help you find what you’re looking for, it helps Google better understand how these things are related.
Carousel: A Touchable List of Answers from Knowledge Graph
There’s also a new carousel view for Knowledge Graph search results. The right sidebar box for Knowledge Graph only shows five related concepts at a time, but now a search for a list of things, like “famous jazz composers,” will produce an illustrated, side-scrolling list of possible answers, which can all be expanded with one tap. It’s a faster and more enjoyable way to dive into search results on the desktop, but it’s also an ideal interface for a tablet.
Gmail Messages in Search Results – A Field Trial
On Wednesday, Google launched a limited field trial of a bold new idea: displaying your Gmail messages in appropriate search results.
That might sound scary, but it’s implemented in a conservative way. It pulls up Gmail messages on the right sidebar if they’re relevant, but the email itself is collapsed. A person over your shoulder won’t be able to read your email unless you click to expand it.
In a publicity event on Wednesday, Sagar Kamdar, director of product management for Universal Search, told reporters, “Gmail is almost the same size as our Web corpus.” Email is a huge store of information that is valuable to users. But for Google to pull email into search results like it does Web results, “Now we need to make it private and secure.”
Collapsing email message views by default is part of that. It’s also controlled by the same toggle switch between personal and global search results as Google+ Search Plus Your World results. Clicking the globe button takes all personal info out of search, and you can permanently disable all personalized search in your Google account settings.
Only if you explicitly tell Google that you’re looking for something of yours, such as by searching for “my flights,” will it show email contents instantly in main search results. It’s all done over SSL, so no one can intercept the information, just like on Gmail itself.
That view of flight info is pretty neat, too. It’s not just about pulling up email messages as search results. Google is actually pulling out the information that’s relevant to you from Gmail and presenting it like a Knowledge Graph answer.
Gmail in search results is a field trial, so you only get it if you want it and sign up for it. You cansign up for the field trial at g.co/searchtrial and see if you like it. It’s not yet available for Google Apps users who have Gmail on their own domains.
Voice Q&A Search In the Google iOS App: What’s A Siri?
The new version of the Google search app for iPhone, iPad and iPod Touch brings to Apple customers the kind of Knowledge Graph-powered, natural language, question-and-answer search Google showed off in Android 4.1 Jelly Bean at Google I/O in June.
This is the most Star Trek of all the announcements; mobile engineering director Scott Huffman bombarded the iPhone with casual questions, and it always instantly nailed the answer, presenting it in a visually attractive and informative format.
Google submitted the app about a week ago. It’s obviously a touchy subject, since Apple’s AI assistant, Siri, is supposed to be the onboard robot that answers people’s questions, and she goes out of her way to avoid using Google to do so. But since Google’s voice search can’t hook in to iOS contacts, phone, messages, reminders, clock or other system-level functions, Siri still has enough advantages that Apple will let Google ship this update.
It’s better than Siri at the Web, though. Let’s get that out of the way. While Siri may improve over time, it just can’t match Google at Web search right now. It can also provide answers any way the Knowledge Graph can, which includes doing math. Siri uses Wolfram Alpha for math and science questions, and the race between Google and Wolfram Alpha will be interesting to watch. But for the casual math and science queries we might ask of our phones, Google is on the ball.
Voice search isn’t the same as Google Now. The predictive suggestions showed off at Google I/O, which send users scheduling reminders from their Google Calendars when searching for a place to eat, for example, are exclusive to Android.
The Destiny of Search
“The destiny of search is to become the Star Trek computer, that perfect, loyal assistant,” SVP of engineering Amit Singhal told the audience in his introduction. We ask questions in natural language, and she answers us.
There are 30 trillion URLs on the Web, Singhal said, and Google crawls 20 billion on the average day. It performs 100 billion searches per month. That represents a staggering amount of information, but it’s a big leap from indexing those Web addresses to actually understanding them. “The next scientific challenge that we face,” Singhal said, “is that we need to understand what’s in that knowledge.”
It’s about more than just organizing the information. If it wants to build the Star Trek computer, Google has to tackle some of the hardest computer science problems there are. It has to master speech recognition, natural language, and, as Singhal said, “We will have to build artificial intelligence. We are clearly not there yet, but we are taking baby steps toward that future.”
The most important change to Google search so far this year was the launch of the Knowledge Graph in May. Google is moving away from simple keyword matching and toward a natural-language understanding of terms and concepts.
By creating a huge database of the things in the world and mapping their relationships, Google is building a way to relate the queries we ask of it to the actual concepts we’re talking about, modeling our language much like we do when speaking to one another. As Knowledge Graph lead PM Emily Moxley put it to ReadWriteWeb in July, “It’s about mapping the real world into something that computers can understand.”
There are 500 million things mapped in the knowledge graph, and 3.5 billion relationships between those things. But Singhal admits, “This is still a baby step towards understanding this world like you or I do.”
“Truly universal search will have information available on the Web and information that’s your information,” Singhal said. Building on Google’s Universal Search, the idea that images, maps, videos and Web pages can all results of the same kinds of searches, it’s time for Google to start bringing users’ own information into the results as well.
The updates announced on Wednesday bring that knowledge to bear in new interfaces designed to make searching Google faster, more natural, and more comprehensive, all while remaining considerate of users’ privacy and devices of choice.