The number of cameras per person on Earth will grow exponentially

This is something I wrote back in 2016 -- per an earlier LinkedIn post, it's an effort to shed some light on why we started OpenSpace in the first place. The original post is here; below is a copy of that post with some minor punctuation type edits. The post hints at one of the trends we are working hard on today -- what we call Spatial AI. We will be sharing more about that idea over the next year or two. Actually, my co-founder Michael Fleischman will be discussing it on a podcast tomorrow (6/5).

Ok here's that old post!

A short post with some pontifications and not a lot of research. You were warned.

Our lives change each time the number of computers per person grows by 10x. In the old days, there were far fewer than 1 computer per person — think the mainframe era. For a while, the ratio was <<<1:1 computer to person. Access to compute was scarce, and mediated by institutions such as universities, governments and large companies.

The PC brought the ratio closer to 1:1 in the 70’s and 80’s, dis-intermediating data, compute and communication. From the 90’s til now, the ratio has grown higher, as people both in the developed and developing world have multiple computers per person that they actually interact with — phone, laptop, sometimes a tablet — and a host of computers hidden in plain sight — in appliances, TVs, cars, everywhere. And of course this is to say nothing of amount of compute each person uses via the cloud.

And new devices continue to appear, creating new categories of digital interactions — from wearables, to voice devices like Amazon’s Echo, to industrial automation and robotics.

I have no idea what the ratio of computers:people is today, but I wouldn’t be surprised if it’s 1000:1.

This ratio has been discussed at length, and it is a useful way to look at how the world and human experience change as technology advances.

I’d also like to suggest a new, related ratio as food for thought: the ratio of cameras per person, over time.

It’s my belief that this ratio is also growing exponentially, and that each order of magnitude increase will usher in big transformations to the human experience.

Some observations.

In the <<<1:1 person to camera era — before easy-to-use cameras pioneered by companies like Kodak — imagery and image capture were mediated by professionals. The 1:1 era began with these Kodak cameras, and then went through a huge change with the advent of cellphone cameras, as connected cameras vastly simplified the capture-to-publish process. (Instagram, ya nailed it.)

So what happens as the ratio goes to 10:1, or 100:1? In fact, why would this growth even happen — what’s the demand for it? With that many cameras out there, there won’t be enough time or people to watch all that content. So what would drive the ratio beyond 1:1?

Increasingly, cameras will be used not just for stills and video to be consumed by humans. Rather, they will produce data to be consumed by machines, and then turned into information of all kinds — humans will never even see the “raw” video.

For example, in the world of drones, you’ll have aerial cameras capturing lots of photos and videos of a construction site. The end output won’t ideally be a list of images or long videos. It’ll be a map, or a 3D model. Or maybe the output has no images at all — it’s just a text report that tells the site manager something like, “Project is going according to spec, except depth of foundation is not adequate. Recommend re-measurement before pouring.”

The human will never see the video. They won’t need to. They won’t want to.

Even today, you can already find sets of cameras ganged together in stereo pair arrangements purely for the purpose of obstacle avoidance (see the Phantom 4 drone) — that video is not for human consumption at all.

My examples are about drones, but the camera explosion won’t be limited to just flying things.

Put simply, cameras are awesome sensors. They can learn about the real world at a distance, unlike inertial sensors like compasses, accelerometers, and gyros. They can capture really high-resolution information.

But there are a few factors gating the next camera explosion, the one that will be all about machines watching video.

The required bits for this to happen include: easy to use pattern recognition techniques that allow machines to interpret vast hordes of data (check, see Tensorflow), powerful computers to run these techniques in the cloud, and maybe at the edge too (check to both), cheap cameras that can be deployed anywhere easily (check), connectivity to get the data to the big computers (check).

The outcome of the coming camera explosion is that the internet will become more and more knowledgeable and aware of the physical world, by directly measuring it, as opposed to guessing what’s going on by looking at what humans are telling it via their tweets, posts, snaps, purchases, and so on.

What will that mean for us?

A lot. Machines that require less active human input, from cars to industrial machines to home appliances. Information workers will orient more toward decision making than “manual” data crunching and massaging. Our notions of privacy will have to change as more of what we do is actively and passively brought online.

I can’t imagine all the possibilities here. But I’m confident this trend is real.