G-Privacy

With all the hoopla over the new Street View mode in Google Maps, I think it’s time to revisit Google’s Mission and issues of privacy. The mission, again, is to organize and deliver the world’s information.

The problem many people have with Street View is that it takes pictures of actual people doing actual things. Is that considered the world’s information? A man peeing by the side of the road. A woman showing us her thong…

But this technology is just a first step, not ultimately meant to show us this snapshot of time, but to give us an idea of the more permanent fixtures of any given point in space. In other words, it’s meant, in the end, to virtualize and mirror the real world — emphasis on the virtual.

Which is why, IMO, this particular technology isn’t quite ready, I’m sad to say. Don’t get me wrong. It’s very impressive, and quite useful too. Just shifting through all of those images is a daunting technological feat. Editing out the billboards is even harder. But it needs to go to the next step — to actually virtualize the full experience — remove the cars, remove the people, remove the particular time of day and weather, and show us only the information we’re looking for.

Google may be forced to do this sooner than later. The invasion of privacy issue alone may make this a big legal hassle — only news organizations get the leeway to use our likenesses for commercial purposes without our permission, and that’s only because it’s ostensibly a public service if we’re part of the news. Movies, for example, must get a release from each person they film on the street. And it’s one thing to archive the web pages I publish. It’s another to become a virtual omni-present paparazzi, an omni-razzi, if you will. Private citizens can and do get more protections from that sort of thing. Why? Because we’re in charge of our own society. And we set our own rules, perhaps not as quickly as companies can, but the people get their way in the end.

From a technical perspective, what this all means is Google will soon be using these billions of images as input to some more advanced algorithm — perhaps what the Stanford Grand Challenge (computer vision) team is helping them with — to build a photo-realistic 3D world, complete with store fronts and road signs and even grass, but free of the extraneous stuff, people, cars, etc.. And at that point, it may even be worth adding virtual people back in — at least those who opt in, and choose to broadcast their avatar’s position, or randomized stand-ins. But until the world is so virtualized, they have a problem beyond the legal one. They can’t easily add virtual cars or people, and the ones presented in the image form too much of a distraction.

This, by the way, is another manifestation of a similar problem in the overhead, aerial and satellite shots. They’re taken from a specific angle, at a specific time of day and weather. Even adding 3D buildings seems strange when you can see a part of the photographed building (as seen from the air) plastered on what should be the street. So the ideal version removes all real shadow and perspective image warping (combining multiple source images, most likely) to show a truly virtualized 3D version of the world that can then be adjusted, moving the sun, adding clouds, or weather, or the projected shadow of a building that doesn’t yet exist.

Once Google can do that — strip out the people and other image artifacts — they can finish the mission of providing, ironically enough, a more accurate and useful snapshot of the real world.

The same thing goes for our search histories and everything else Google stores. There are some (opt-in) reasons for Google to keep this data, to provide a few personalized services. But for the majority of people, the answer is that Google must do more to anonymize the data — not after one or two years of storage, but immediately, if they are going to retain the information at all.

It’s fine for Google to present an aggregate view of the web as it is, including all of it’s virtualized people. But Google should not become a de-facto version of the NSA, or Star Magazine, or a peeping eye in the sky, available about anyone to anyone who asks. The ultimate end of this technology is not (I hope) a real-time view of every spot on Earth, but a virtualized version that provides the answers to the questions what? where? and –for people who opt-in– who?