1 Picture = 1,000 Words. 1 Picture x Earth’s Surface x 365 x 2 = Knowledge

by swtouw

Look at an overhead picture of anywhere at any time.

Sit back and think about the possibilities for a second.

Thinking about going to the gym?  Check the parking lot to see how crowded it is first.

Wondering what side of the road the accident ahead of you is on?  Look at the image of the road ahead.

Trying to find where your buddies are tailgating?  Look for their car before you drive over.

Long ago, in an old job I can’t talk about, high resolution pictures taken from outer space was magic stuff…

It wasn’t that long ago, though.

Now, we take it for granted that we can look at an image of our house on Google Maps, or even a street view of the address for your meeting tomorrow.

Typically these satellite images are years old, maybe months if you’re lucky.  Google buys them from a commercial satellite company like DigitalGlobe (another previous employer) and fuses them into their globe for your enjoyment – the acquisition, transfer, and fusion takes time, so you’ll never see an image hours old.  Hi my house last summer!

Until now.

Well, not now, but really soon.

Satellites are the easy part, at least it seems like it is now.  Check out the latest Google acquisition, Skybox.  What does Google want with a Satellite company? — more like, what doesn’t Google want with a Satellite company!?

In fact, they aren’t a satellite company at all, and their founders agree.  They believe they are creators of knowledge – from data that happens to be collected across the globe, twice a day by 2016.  Will you be able to get it in minutes or on-demand by 2020?  That doesn’t seem to be crazy talk anymore.

You want to talk about big data – that’s a hell of a lot of pixels to store and index every day.  This is the kind of massive data dreams are made of.  Not only must Skybox and Google figure out how to index this data effectively for rapid retrieval (wonder if Google will do that well?), but they must be able to analyze it.

This isn’t some college grad on a light table – no, to create knowledge, the image data must be analyzed in automated ways, at scale.

Talks of a developer API for this data has been rumored.  Maybe that means on-demand query of image by location at first.  But let your imagination go crazy for a second — what about running algorithms over the same orthorectified area over the past year to create time series statistics.  That starts getting really complex for the API provider, a framework that not only provides access to that data, but scales your algorithms to process the data.  Dream job?

The future should be about providing not only data services, but data-analytic services.  Developer APIs should be about the logic of what they are trying accomplish, only.  Let the API service provider worry about data access, scaling of jobs, and indexing of results.

So, yeah, Google could keep this to themselves and build their own hedge fund based on this new source of knowledge no one else has (and also build some damn good maps).  Or, they could open up a platform API that allows retrieval and analysis of this data to build knowledge I can’t even imagine.

I hope they do the latter.

Quick update (7/30/2014):  Just saw Will Marshall of Planet Labs’ exceptional talk at OSCON, very exciting stuff…looks like this is going to happen, here’s the video.