Wednesday, December 27, 2006

Crowd Wisdom vs. Google's Genius

Can the wisdom of crowds trump the genius of Google? The founder of Web encyclopedia Wikipedia believes it not only can, but it will. Jimmy Wales, the man behind Wikipedia, which ranks among the top 15 online destinations worldwide, plans to launch a new search engine in the first three months of 2007. He believes the project, which is called Wikiasari, could someday overtake Google as the leader of Web search.

Like Wikipedia, the new search engine will rely on the support of a volunteer community of users. The idea is that Web surfers and programmers will be able to bring their collective intelligence to bear, to fine-tune search results and make the experience more effective for everyone. "If you search in Google, a lot of the results are very, very good and a lot of the results are very, very bad," says Wales. What that shows, Wales says, is that mathematical formulas alone do not produce consistently relevant results. "Human intelligence is still a very important part of the process," he says.

People can contribute to Wikiasari in one of two ways. The first is by enabling ordinary computer users to rerank search results. When a user performs a search on Wikiasari, the engine will return results based on a formula akin to Google's own Page-Rank system, which determines relevance by counting the number of times other Web pages link to a specific page, among other things. Unlike Google, however, users will then be able to reorder the results based on which links they find most useful by selecting an edit function. Wikiasari's servers will then store the new results along with the original query. When the same query is made in the future, Wikiasari will return the results in the order saved by most users.

"Google-Killing" Potential

Web users with programming knowledge have a second way to contribute. Wikiasari's technology is based on Apache's open-source Web search software Lucene and Nutch, and Wales plans to unveil all the company's computer code to the outside world. This kind of open-source development is in sharp contrast to the approach of the leading search engines, which do not release their search ranking formulas. Yet Wales contends that his open approach will ultimately prevail, because anyone any place in the world can weigh in with tweaks to Wikiasari's code to help return more relevant results.

Google has proven to be a fearsome competitor, however, against some of the most powerful companies in technology. In recent years, Google has increased its lead in search over Yahoo! and Microsoft's MSN, despite vows from both companies to catch up. Google controlled 49.5% of the searches in December, up from 43% two years earlier, according to Nielsen//NetRatings. Meanwhile, Yahoo has 24.3% of searches and MSN, the third highest ranking search engine, controls just 8.2% of searches. Microsoft's share of the search market has slipped from 14% two years ago, despite a technology revamp and a multimillion-dollar ad campaign. If the searches that Google performs for Time Warner's AOL and News Corp.'s MySpace are added to the equation, its lead is even greater.

Though the Wikiasari project is scheduled to debut in the first quarter of next year, Wales suspects it will take roughly three years of user input before it has enough information to become a real competitor to the top search engines. "I wish I could write a Google-killing project in three months, but it is going to take a little bit longer than that," says Wales. "We will have something up in '07, but it won't be very good or interesting. It will be a starting point for people to play with."

This Time, It's Profitable?

Wales bases his three-year timetable on the amount of time Wikipedia took to really take off. That project launched in 2001 with about 100 articles. It now has more than 6 million articles in a variety of languages, according to Wikipedia's own Wikipedia entry. Though the trustworthiness of its user-supplied content is hotly debated, at least one study showed it is now nearly as accurate as other encyclopedias and frequently used by scientists.

However, there is a key difference between Wikiasari and Wikipedia. While Wikipedia is a nonprofit, Wikiasari is a for-profit venture by Wales' company Wikia Inc. Wales has received some high-powered backing for his venture—$4 million in funding from Bessemer Venture Partners and Omidyar Network. Bessemer was one of the original investors in Internet phone phenom Skype, and Omidyar Network is the investment vehicle of Pierre Omidyar, founder of eBay. E-commerce giant Amazon also put money into the company earlier this month. (Contrary to one published report, Amazon is contributing only money, not technology.)

Wikiasari will serve ads along the right side of the search results page, says Wales. Like Google, the ads will be related to queries and clearly marked as sponsored links. The presence of ads could make the kind of users who contribute to Wikipedia respond differently to Wikiasari. After all, contributing to Wikipedia is seen as something of a selfless act in support of a nonprofit that exists solely to benefit the Web community. Supplying information to Wikiasari, on the other hand, will benefit Wikia and its advertisers.

Other Community Experiments

Wales doesn't think people will care. As proof, he uses the example of Red Hat, a public, for-profit company that relies on Linux's free, open-source operating system and user collaboration. The company made $105.8 million in the third quarter of this year, up 45% year over year. "It's all about free licensing and sharing your work with others," says Wales. "They don't care about people making money; they care about people taking their work and locking it up."

Even if people do contribute, that doesn't mean Wikiasari will eventually become leader of the pack, or even one of the leading players. The major search engines know that links alone are not infallible indicators of the best results. Both Yahoo and Google have experimented with ways to involve the community to refine their own search products. Yahoo, for example, has its Answers property that relies on volunteer users to supply information related to specific queries. Similarly, Google has allowed people to vote on results by clicking on a smiley face on its toolbar for good results and on a frowning face for bad ones. Google shut down its own answers service, with real people responding to questions, in November.

Search's Next Level

Other companies have tried to incorporate human intelligence in search results. StumbleUpon, for example, has a toolbar that shows users which pages in Google, Yahoo, or any other engine's search results have been rated favorably by its community based on the users' interest. The 3-and-a-half-year-old company, funded in part by Ram Shriram and Rajeev Motwani, two of the investors who initially backed Google, has grown from 500,000 users to 1.65 million users in the past year alone. "Algorithmic search can only get you so far," says Dave Feller, StumbleUpon's vice-president of marketing. "The info from other people can get you to the next level."

StumbleUpon's ambitions may suggest what the future holds for Wikiasari. Feller says that the company would be open to partnering with a Google or Yahoo. With Google's $140 billion market cap, it could even buy such a toolbar company to enhance its own results. Wales may have dreams of taking on Google with his new Wikiasari project. But he may end up developing a search engine that will make Google—or one of its rivals—that much more effective.

Holahan is a writer for BusinessWeek.com in New York.

Wikinomics: How Mass Collaboration Changes Everything

Wine investing: buyers, cellars and liquid assets