Mountain view, Calif. — In early 2008, in the early days of the iPhone era, Google engineers began noticing something unusual in the search engine’s logs. Owners of these new phones were doing a huge number of Web searches.
Peter DaSilva for The New York Times
A photo can be used for Web searching with the Google Goggles smartphone app.
But there was a problem: searching on a phone was less than ideal. It was hard to type on small screens. And most irritating for Google, which brags about its speed on every page of search results, was that Web pages were slow to load on phones.
So Google started a project it code-named Grand Prix. In six weeks, engineers revamped mobile searching and hatched plans for new ways to search on the go, by talking or taking photos instead of typing.
The stakes were high. Mobile phones could be a huge new market for Google. Or they could provide an opening for a competitor to pounce, or obviate the need for a search engine altogether. If people on phones could go straight to apps for information, why Google anything?
Today, Google says mobile searches are growing as quickly as Web searches were at the same stage in the company’s early days, and they are up sixfold in the last two years. Google has a market share of 97 percent for mobile searches, according to StatCounter, which tracks Web use.
Now that it dominates the field, Google is throwing its burly computing power and heaps of data at new problems specific to mobile phones — like translating phone calls on the fly and recognizing photos of things like plants and items of clothing.
Peter DaSilva for The New York Times
Scott Huffman, left, and Amit Singhal of Google. With voice search, Google learned it had trained people to use keywords.
Scott Huffman, left, and Amit Singhal of Google. With voice search, Google learned it had trained people to use keywords.
“I feel like a parent the second time around feels,” said Amit Singhal, a Google fellow who works on search. “You saw your first child grow at an amazing pace, and here we are with our second child, mobile, growing at the same pace and showing the same signs.”
Google has been slow to seize some newer Web business opportunities, most notably social networking. Investors have criticized the company for dragging its feet when it comes to figuring out how to make money in new fields.
But mobile is an exception. Last year, Eric E. Schmidt, then the company’s chief executive, said Google’s philosophy was “mobile first,” meaning it would build products for phones at the same time as versions for PCs.
“This is the place that Google is essentially betting its future on,” said Karim Temsamani, Google’s head of mobile advertising, a role created in September.
Still, Google has not consistently followed the mobile-first mantra, and some analysts, including Colin W. Gillis of BGC Partners, say it has not moved quickly enough to create new mobile products or ads.
“They’ve done a really good job of positioning themselves so they can’t get boxed out of the market,” Mr. Gillis said. “Now they just need to deliver some innovation. Let’s wring some revenue out of this platform.”
Google said in October that mobile ads were on track to generate $1 billion in revenue in the coming year. Mobile users can call a business from within a Google ad or receive coupons for nearby stores. They can take cellphone photos of movie posters to pull up a trailer. With new technologies like near-field communication, advertisers could reward customers with loyalty gifts for walking into stores, Mr. Temsamani said.
But because mobile ads generally sell for less than half the price of Web ads, Mr. Gillis said, “there’s just not a lot of profit left over.” Though Google makes Android software for phones, it does not make money from it directly because it gives it away to phone makers. Meanwhile, Apple makes money from its devices and from what appears on their screens, including its own ad network.
Still, the company’s approach to the mobile market is classic Google: take problems that computer scientists have been working on for decades, throw huge amounts of data and computing power at them and assume that if the resulting product is useful to people, it will eventually make money.
People can now snap photos of landmarks or wine labels to search for them using Google Goggles, speak to their phones using voice search and, on Android phones, translate spoken conversations between English and Spanish.
“We as an academic community would have figured this out, but we wouldn’t have been able to set it up on this kind of scale,” said Alexei A. Efros, an associate professor in computer science and robotics at Carnegie Mellon, referring to these kinds of technological feats. “That’s really the great thing about Google, the fact that it can do it on such a humongous scale and actually make it useful to the general public.”
Google trained its computers to learn spoken language based on troves of voice recordings. “Even if you’re from Brooklyn and you drop all your R’s when you park your car, it’s heard plenty of people from Brooklyn and it can do well,” said Mike Cohen, head of Google’s speech technology team.
At first, Google engineers thought people would talk to its voice search service as if they were talking to a person — “you know, it’s my anniversary, and I’d love to take my wife somewhere really romantic to eat, do you have any ideas?” — so it taught the service to filter out unnecessary words. But it turned out that Google had already trained people into thinking in keywords, so they knew to search “romantic restaurants” even when speaking instead of typing.
Goggles, the visual search tool, recognizes things that have strong visual textures, like a bar code, book cover or landmark. But it often can’t distinguish between a black cat and a black chair, for instance, or recognize food or plants, though Google is working with botanists to teach its machines the secrets of leaf-spotting. Google already has the capability to recognize faces, so people could theoretically snap a photo of a blind date and pull up an online profile, but it is not yet using that technology because it is still working out the privacy implications.
People can also snap a photo to translate a menu in a foreign country, and speak English to hear the Spanish translation. Someday Google hopes to be able to translate both sides of a phone conversation as it happens, said Franz Och, head of Google’s machine translation group.
Though the search results Google spits out might seem the same on phones as on computers, there are some behind-the-scenes differences.
For example, certain search results are ranked differently, with location factored in. Search for Wal-Mart on a computer and Google suspects you are probably looking for the e-commerce site or job openings. Search on a phone and Google assumes you are looking for the nearest store. Other search tools were built specifically for phones. Search for weather or stock prices and Google shows a scale, movable with a finger, to see results for different times.
Google says mobile search is not stealing time from computer searches. Instead, mobile searches spike during the lunch hour and evenings, when people are away from their computers. And while mobile users do search for simple things like weather and train times, engineers have been surprised at how many people also ask more complicated questions about business and politics.
“Mobile search is definitely going to surpass desktop search,” said Scott B. Huffman, who works on mobile search at Google and leads its search evaluation team. “The lines will pass, and I think they’ll pass before anyone thought they would.”
© 2011nytimes.com
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.