We just had a quick hands-on of Facebook's new "Graph Search" at one of the demo stations here at Facebook's press event. Unfortunately we weren't whitelisted at the time so we couldn't use any of our own data (update: we did a bit later -- see below) but we now have a decent idea of how it all works. Using the four "pillars" of Graph Search -- people, photos, places and interests -- we looked up terms like "My friends who like Star Wars and Star Trek," "Restaurants in San Francisco that my friends like," "Photos of hiking in national parks" and managed to get results every time. There's a side bar of drop-downs to refine the search further -- you can narrow the results by gender, categories, interests, etc. Overall the UI is clean and intuitive; those who are already familiar with Facebook should find no problem navigating the interface. The results are listed vertically in a pretty straightforward manner, as seen in the photo above.
In addition to playing with the Facebook's newest function, we briefly spoke with Tom Stocky, former Googler and director of product, about the challenges of creating Graph Search. He told us that two of the biggest hurdles faced by his team were dealing with the massive amount of data and incorporating natural language search. To solve the second problem, the company brought in computational linguists to work with Facebooks's engineers, and for the most part, it seems Facebook succeeded in its natural language efforts.