I signed up for the Facebook Graph Search beta a couple weeks ago. The invite came the other day and I eagerly clicked through to activate it. Immediately, a new search box appeared at the top of my newsfeed.
I started typing right away, and... There it was! Page after gridded page of people who were just like me - they liked Bacon and Parker Guitars! If I wanted to, I could narrow the search by increasing the number of interests, or selecting geography, or specifying relationship status, connections to other friends... Any number of ways to pivot the data. You can also flip it and look for things using your friends as the pivot. I was able to look for music my single friends liked, and asian restaurants liked by my Chicago friends.
I was having a blast.
Now, for the downside... The search seems to have some teeth and can access preference information that you didn't really think about being public. As a result, some fairly powerful inferences can be made, as illustrated by @tomscott on his tumblr feed, Actual Facebook Graph Searches. This is bad news if you're a Jew who likes Bacon and don't want your mom to find out. It's dangerous because it can illustrate an association between you and your less savory acquaintances.
Fortunately, a search on my friends who like Racism turned up an empty set.
So, like any powerful thing, it can do great good or great harm. However, the powerful interest for me is that, unlike traditional search, it is not just searching entities. Even Google's powerful index is making content as an entity searchable by looking at it's relationships with other content entities. What Graph Search does, is make the relationships themselves searchable.
The technology is still new and, in many ways, still obviously shows it's youth. However, as the technology and indexing improves, the ability to search for relationships between things as a way to turn up interesting slices of a large data set will make Graph Search a powerful tool for anyone looking for great insights.