Former Googlers Lars Rasmussen (left) and Tom Stocky have created a new search engine for Facebook leveraging data that Google can’t touch. Photo: Joe Pugliese
Beast had a birthday last week. The First Dog of social networking — live-in companion to Mark Zuckerberg and his bride, Priscilla Chan — turned two. The proud owners baked a cake for the Hungarian sheepdog and decided to throw an impromptu party. Naturally, when it came time to compile the guest list, the couple turned to Facebook, the $67 billion company that Zuckerberg founded in his dorm room nine years ago.
To date, sorting through your Facebook friends could be a frustrating task. Although the site has a search bar, there has been no easy way to quickly cull contacts based on specific criteria. But Zuckerberg was testing a major new feature that Facebook would announce on Jan. 15 — one that promises to transform its user experience, threaten its competitors, and torment privacy activists. It’s called Graph Search, and it will eventually allow a billion people to dive into the vast trove of stored information about them and their network of friends. In Zuckerberg’s case, it allowed him to type “Friends of Priscilla and me who live around Palo Alto” and promptly receive a list of potential celebrants. “We invited five people over who were obvious dog lovers,” he says.
For years now, Facebook watchers have wondered when the company would unleash the potential of itsunderpowered search bar. (Nobody has feared this day more than Google, which suddenly faces a competitor able to index tons of data that Google’s own search engine can’t access.) They have also wondered how a Facebook search product might work. Now we know. Graph Search is fundamentally different from web search. Instead of a Google-like effort to help users find answers from a stitched-together corpus of all the world’s information, Facebook is helping them tap its vast, monolithic database to make better use of their “social graph,” the term Zuckerberg uses to describe the network of one’s relationships with friends, acquaintances, favorite celebrities, and preferred brands.
In the weeks leading up to the launch, Facebook executives were still trying to come up with a name for the new product. They were hoping to stay away from the word “search,” to distinguish it from web search. (Only a few days before the launch, one Facebook executive slipped and referred to it as “browse.”) But after hours of contortionism, they relented; nothing topped Graph Search. “It’s descriptive — it’s search,” Zuckerberg says. “And the graph is a big thing.” The idea is that Facebook’s new offering will be able to extract meaning from the social graph in much the same way that Google’s original search unearthed the hidden treasures of the web. “People use search engines to answer questions,” Zuckerberg says. “But we can answer a set of questions that no one else can really answer. All those other services are indexing primarily public information, and stuff in Facebook isn’t out there in the world — it’s stuff that people share. There’s no real way to cut through the contents of what people are sharing, to fulfill big human needs about discovery, to find people you wouldn’t otherwise be connected with. And we thought we should do something about that. We’re the only service in the world that can do that.”
The result is surprisingly compelling. The mark of a transformative product is that it gets you to do more of something that you wouldn’t think to do on your own. Thanks to Graph Search, people will almost certainly use Facebook in entirely new ways: to seek out dates, recruit for job openings, find buddies to go out with on short notice, and look for new restaurants and other businesses. Most strikingly, it expands Facebook’s core mission — not just obsessively connecting users with people they already know, but becoming a vehicle of discovery.
Zuckerberg says that this is in fact a return to the company’s roots. “When I first made Facebook, we actually offered some functionality that was like this but only for your college,” he says. “Facebook then was arguably as much for meeting new people around you and exploring your community as it was for keeping in touch with the people you already knew. But it was such a hard problem to do it for more than a few thousand people at a time. We transitioned from connecting with whoever you wanted to primarily staying with people you already knew. But Graph Search is like the grown-up version of that discovery aspect. Exploring your community is a core human need, and this is the first big step we’re taking in that direction.”
The first of many steps, that is. Graph Search will be improved based on how people actually use it. So Facebook plans a slow introduction, limiting the initial rollout to a small number of users. Zuckerberg’s expectation is that by the time it becomes available to millions it will be considerably improved.
For example, he thinks he can make it easier to find invitees for a canine birthday party. “We don’t have the ‘who has dogs’ field yet,” Zuckerberg says.
Team Graph Search: Stocky, Rasmussen, and their crew of about 70 at Facebook HQ in Menlo Park, California.Photo: Joe Pugliese
Graph Search got its start in the spring of 2011, when Zuckerberg asked Lars Rasmussen to join him on a walk. Rasmussen, 44, had joined Facebook the previous year, an eyebrow-raising defector from Google. The Denmark-born engineer’s career had been distinguished by two projects, one a major win and the other a legendary fail. The first was Google Maps, which began as a small company based on a brainstorm by Rasmussen’s brother. Google bought the company in 2004 and expanded it into a landmark product so central to our digital existence that users almost rioted when Apple dropped it as the iPhone’s default mapping app. The second product was Wave, a complex system that mashed up conferencing, e-mail, and messaging. Rasmussen and his team got Google to devote $25 million and 60 engineers to the project, which he introduced in 2009 at a Google developers’ conference. His 80-minute product demo won a standing ovation. But Google pulled the plug in 2010 — Wave turned out to be too confusing to win over many users, marking what Rasmussen later called “the most painful and spectacular failure of my life.” A few months later, Rasmussen left Google and joined Facebook.
Now, walking with his young boss, Rasmussen was offered the chance to roll the dice once more. Facebook, Zuckerberg said, had a unique opportunity to deliver fantastic value with a different kind of search — detailed and targeted dives into its huge, structured database. Rasmussen was sold. “We could build a compelling new pillar,” he remembers thinking.
Rasmussen joined Facebook’s existing search team. The company already had truckloads of information, but it was hard for users to access. Who are my friends in New York City? What books are my friends reading? Is there anyone nearby who loves Wilco? What’s an Italian restaurant that people really like? The new search product would answer such queries. But Rasmussen’s team faced a tough quandary: whether to focus on the most popular kinds of questions — or take on the tougher challenge of building a smarter search engine that would let users ask Facebook pretty much anything.
That summer, Rasmussen, armed with a crude demo, met with Zuckerberg in a glass-walled area of Facebook’s then-headquarters called the Aquarium. The proto-search engine was limited to predetermined queries. But the engineer laid out a more expansive vision — the more ambitious approach that would allow the engine to process virtually any query. He talked of a search engine that could answer requests like “show me pictures of my friends and me visiting California in 2010.”
Zuckerberg now says that when he saw the demo, he thought that Rasmussen’s approach was dead right — but probably impossible. “No chance,” he says now, thinking back to his reaction. “You could type in anything you wanted and it would be the title of a new page with the content just magically laid out. No one’s gotten natural language to work like this. And then to actually be able to index all the stuff. There’s more than a trillion connections on Facebook! Building up the infrastructure to index all of it and be able to cut it in any way is a monumental technology challenge.”
Nonetheless, Zuckerberg enthusiastically expressed how sweet it would be if such an approach were actually implemented. In Silicon Valley–speak, that’s a direct order.
That summer, Rasmussen got a co-leader for his product: another former Googler named Tom Stocky, an MIT grad who had worked on various teams since joining Google in 2005. (This bears repeating: Facebook’s search product was led by two ex-Googlers.) His previous post there was as a director of product who worked on travel search — which made him a compelling target for Facebook’s search group. “They told me the vision — let’s make everything searchable and discoverable.”
For more than a year, Rasmussen and Stocky met Zuckerberg every Friday at noon to update him on their progress. Eventually 50 engineers would work on the project, including two linguists to help the engine understand people’s queries.
As the true scope of the search engine emerged, it became clear that Graph Search would require a full Facebook makeover. To encourage people to write more detailed queries, Facebook made the search bar bigger, essentially dominating a wide blue banner toward the top of the page; various icons were shoved to the margins. Most strikingly, the name of the company itself was dropped from the home page, replaced by a single stylized F. In other words, Graph Search was important enough to bump the word Facebook from Facebook. “When I first joined the team, I was a little skeptical: Is search really going to be the quintessential part of Facebook in the future?” says Keith Peiris, a product manager on the search team. “But we quickly realized that this was inevitable and would make Facebook stronger.”
The big adjustment was understanding that some of the rules of Graph Search were dramatically different from web search — and that part of the team’s task would be “diseducating” users. Good web search results can be had with very few, relatively vague keywords. But Graph Search works better the more specific and complex the request. To tease out those more complicated queries, Facebook makes guesses (or “type-aheads,” similar to Google’s auto-complete) as to what you might be looking for. Type in “New York” and it might ask if you want to find “friends from New York,” “restaurants people visit in New York,” or “things liked by people in New York.” The more complicated the query, the more precise the answer. Type “what restaurants in San Francisco are visited by my friends who like Homeland?” and you’ll probably get a valid result. “We really want people to unlearn the model of using the three vaguest words possible and instead actually say what they want,” says Peiris.
Now it’s time for Facebook to see for itself what users really want. Rasmussen says he has no idea what will happen, especially after experiencing both the ecstasy of Google Maps and the agony of the ill-fated Wave. “But I was nervous with the first product and not with the second,” he says. “So the fact that I’m nervous as hell is a good sign.”
In early December, Zuckerberg gave me an early look at Graph Search. He warned me that the product was still rough — at that point it didn’t even have a name. We met, along with Rasmussen and Stocky, in his conference room on Facebook’s new Menlo Park campus, formerly the home of Sun Microsystems. The ground-floor room was walled by glass; like its predecessor in the former Palo Alto headquarters, it’s dubbed the Aquarium. Zuckerberg, wearing his iconic hoodie, started the meeting seated; but as he spoke, his enthusiasm apparently compelled him to move around the room, occasionally bouncing a soccer ball for emphasis.
Rasmussen typed in “photos of my family,” and instantly a grid of pictures filled a display screen. “This to you will look like a bunch of silly blond people — I’m from Denmark,” he said. “But to me, this is just an awesome experience.” Then he asked for friends in New York, and it gave him the list.
“One of my favorite queries is recruiting,” Zuckerberg said. “Let’s say we’re trying to find engineers at Google who are friends of engineers at Facebook.” He typed in the query and found, not surprisingly, that there were lots of people who met those criteria. Each one was represented by a little rectangle of information — their profile photo, along with snippets of key information like where they went to school, where they live, and the names of the mutual friends. “It’s like Facebook is this big database and you’re doing a lookup on the results that match,” Stocky said.
“The good thing is that there’s people at the end of these connections,” Zuckerberg said. “You can find the right people or content page and then send a message.”
Rasmussen jumped back in. “And suppose I want a job at Pinterest — which I don’t, for the record — and I want someone to introduce me there,” he said. “I can search for my friends who are friends with Pinterest employees.”
read more at http://www.wired.com/business/2013/01/the-inside-story-of-graph-search-facebooks-weapon-to-challenge-google/all/
Δεν υπάρχουν σχόλια:
Δημοσίευση σχολίου