From a long while I’ve wanted to build a camera to watch over the little road next to our house. There is a smaller road next to the main road. The main road leads to the motorway and traffic builds up most mornings. Long queues of single occupancy vehicles. I imagine the whole traffic jam of drivers in a bus instead. It would take up much less room. Or on a small herd of bikes. That would require some form of cooperation. Instead, all of them get into their cars each morning and wish there were less vehicles around so that they could have an open road ahead of them. Some of them choose to jump the queue by speeding down the little road in front of house. The road on which my kids bike to school or play. Most of the drivers speeding down the road are not thinking of my kids but are just frustrated with the jam and their heart leaps for joy as they find a way of jumping 10 places. At that moment they are probably not aware of the fact that lane changing is one of the core reasons for jams in the first place.

Almost every time I see this happen, I wish I had my own personal speed camera watching over the road in front of my house. I discussed this a couple of times with some friends and I was told in no uncertain terms that this is completely illegal in the Netherlands. You are not allowed to take pictures of the highway, which is formally owned by the Crown. Or some such.

A year or two back, I talked with Andrew Farah, the founder of Density.io. The company has built a 3d printed sensor which, when attached to a doorway, counts how many people go in or out. It does that accurately, cheaply, in real time and most importantly anonymously. The images it captures contain no personally identifiable data. It sends these to the cloud where deep learning translates this very noisy data into in’s and out’s and in turn, how many people are in a area at any point in time. Companies can subscribe to an API which gives them a count of people in their rooms, floors or buildings. By default this busy-ness information is public, so directories and other aggregators can display how busy a business is at any point in time.

Businesses normally publish their place and contact details or “where”: name, address, telephone, email and maybe URL; and some advertising information or “what”: keywords, logo, photos, products, services and pricing. Few business profiles or websites contain much more. If you were to add more, time or “when” is incredibly useful: when you’re open, when different parts of the business are open, e.g. the bar opens until 1am but the kitchen closes at 10.30pm. Sometimes businesses allow their customers and other users to write about them, acknowledging that to be trustworthy it helps if their customers say similar things about them as they say about themselves. Businesses only publish advertising, and user endorsements and reviews only publish opinions, but by combining the two an interested prospective customer can triangulate to the truth.

Within the time a business is open, when to go is one of the most asked questions and one with the least data. When is my favourite bar the busiest? Which club is heaving right now? When is the gym empty? Which co-working space location should I go to to have the most chance of working in peace? Density provides accurate data to answer all these questions and more. In addition, it adds a new data source that is not what the business claims nor what its customers say they’ve experienced.

Last year, I dropped in at The Ground, an awesome tech start-up space in Malmö, full of interesting businesses, each building something scalable. I’m an investor in Zook.ai, based there, and it’s one of my favourite spaces. Unfortunately the commute means that I don’t get to go there as often as I’d like. I met some of the guys from Mod.cam there. The Mod.cam is like a little smartphone which sees around it but its output is a JSON stream of anonymous information. It can generate counts of people, genders, ages and recognises some facial expressions and even work out the paths people take around a shop, but it never stores personally identifiable information.

The applications for anonymous perception are enormous. Currently, humans or images are always needed somewhere in the mix to watch over something. Either at the start, e.g. a patrolling police agent, or by aggregating video feeds into a single human, e.g. a security guard. Anonymous perception can be watchful without ever watching. It can lay dormant all day only to spit out a single answer. I’d like to set up a sensor like this to watch over my street and take pictures of the dog owners and their dogs who take a dump on it, tracks them back to where they go indoors then drone delivers the excrement to their door (or maybe that’s not so anonymous any more). Or a sensor which tells me which distribution of clothing types are most representative of a rich neighbourhood and which of a poor neighbourhood. Or a sensor which locks thieves into my garden when it senses someone it doesn’t know is attempting to remove one of the family’s bikes. Or – a sensor to watch over my road and deliver a JSON feed of vehicle number plates, datetimes and speeds.

Would the authorities be able to trust such a observerless observation? I could imagine a world where the sensor could prove the code it was running was the same as was in the some version of an open source repo, and it would add each of its results to a blockchain of number plates, locations and speeds, stored across each of its brethren of sensors.

Anonymous perception is part of what Evan Nisselson of LDV calls the Internet of Eyes. It requires a shift in language to move from thinking and talking about cameras as being inextricably linked to photos and videos and realise them as instead being visual sensors. As Benedict Evans of a16z says “When we can turn images into data, we’ll find lots of sets of images that we never really thought of as data before, and lots of problems that didn’t look like image recognition problems.”