How Will Google Build Its Serendipity Engine?
Google CEO Eric Schmidt wowed the crowd at TechCrunch Disrupt Sept. 28 with talk of autonomous search and serendipity engines that deliver search content to users' mobile phones without the user having to do anything but walk down the street.
What Schmidt didn't say was how Google would build its serendipity engine.
Given Google's penchant for leveraging algorithmic search, we can logically assume these results will be auto-generated by the sprawling Google search engine, proving a major efficiency boost over Google Instant predictive search, which now provides a major efficiency boost over traditional type-query-hit-enter search.
You get the idea, but I'm not sure that's the best route, if only because Facebook and a slew of startups are setting examples of how people -- not the machine -- are helping people find the information they seek. Let's start with Facebook.
A few days later, Facebook CTO Bret Taylor announced that 2 million Websites have added the Like button and Facebook's other new social plug-ins only since April, or about five months ago.
The Like button lets a user share publishers' content with friends on Facebook. When a user clicks a Facebook Like button on a Website, the publisher will gain a link from the user's profile, the ability to publish to the user's News Feed, inclusion in search on Facebook and analytics.
Facebook partners that feature the Like buttons push to Facebook information about items their visitors liked--for example, bands that users liked from Pandora, local businesses such as restaurants from Yelp and movies from IMDB.com. Information about objects users click on will appear in users' profiles as items they endorse.
Likers, as Facebook called them, are nothing if not prolific. Facebook claimed the average "liker" has 2.4 times the amount of friends than that of a typical Facebook user. They also click on 5.3 tmes more links to external sites than the typical Facebook user.
At the current pace, Facebook may rack up 5 million Websites a year, allowing its Like button to go viral across the Web. Facebook CEO Mark Zuckerberg ultimately expects 1 billion Like buttons to propagate across the Web.
So the Like button is its own recommendation engine, not machine-generated, with real people recommending content. Facebook has proven that it works.
Moreover, while Schmidt proposed search results piped straight to users unfettered, the Like button requires users to take action -- clicking the Like button -- which leads to a cascade of content population on Facebook Web pages and more information for publishers about users.
Facebook Like buttons are just one avenue of personalized recommendations. Startups such as GetGlue, Hunch and My6sense provide new outlets for recommendation.
My6sense streams content from users' RSS feeds and social streams, mining links from Google Reader and Google Buzz for the iPhone and Android applications.
These are all socially leveraged applications and each company, especially Facebook, has a unique handle on how to serve users.
Given Google's new emphasis on social apps to battle Facebook, what form will Google's recommendation engine take in the future? Will it be social, as in content fed and/or curated by users' friends and contacts? Or will it be math-based results and info?
I'm inclined to think Google has its Aardvark team working hard on this challenge. I look forward to what Google offers.