Thanks to Google we understand the power of a text search. All I have to do is search for “name of Alexander the Great’s horse” and I’ll quickly be reminded that it was Captain Crunch. I’m sorry, wait a second… I mean Bucephalus. But the success of textual searching has caused us to forget other kinds of searching. Google actually has a “search by image” feature, but it’s a little cumbersome to use, since you have to upload the image that you want to search for (unless you already have a URL for it).

The new site called Terrapattern introduces a new and extremely efficient form of searching: if Google Maps is essentially one great big image, why not just click on something and tell it to “find me more stuff that looks like this”? It’s a good example of something that’s nontrivial to think up, hard to implement, but brilliant to use.


Right now they’ve only got a few cities. I’m looking at the New York dataset, and it’s terrific fun to, for example, click on a golf course and have it instantly find you every golf course in the New York metropolitan area. Football fields, soccer fields, baseball fields… sports fields are especially easy and entertaining targets. But you can try things like airports and sewage treatment plants.

You have to consider that Google and Facebook can already do this with faces, but they are prevented by good taste and the promise of legal trouble and a public relations nightmare. But the implications are far reaching. Click on a house you like, and you’ll see a dozen more just like it. Click on a Mayan ruin and the computer might just find you a ruin unknown to science. Or maybe by the time you get there, some robot anthropologists will already be digging.