What if AI could make our cities smarter and more private?

There are two very different versions of the AI-powered city coming into view. We need to choose which one we want.

The first is extraordinarily convenient. You pay for public transport with your face. Cameras understand what is happening across the city and help things run more efficiently. Artificial intelligence helps police streets, manage transport and deliver public services.

The second is a potential surveillance state.

A recent New York Times report describes a city transformed by technology: crime has fallen dramatically from the levels of 25 years ago and digital infrastructure has made everyday life safer and easier. But this city is Moscow and it is arguably a police state city with roughly 300,000 CCTV cameras, linked to biometric databases capable of identifying and flagging individuals. The same infrastructure that allows a citizen to live a frictionless life can be used to allow passengers to move seamlessly through a metro gate or can be used by the state against the individual later.

That is not an argument against smart cities. I have spent much of my career developing artificial intelligence that understands what is happening in the physical world and with the goal of making cities work better. I have come to believe that the fundamental question isn’t how intelligent we can make our cities; it is about what we make them intelligent about and how we do this.

Not all sensors are about surveillance. For smart cities, when it comes to maximising curb-side value, traffic efficiency, school zone safety during pick-ups and drop-offs, and noise pollution, sensors are about understanding how an environment operates. It is not about the individual in that environment, per se.

This distinction matters because if you build the technological solution with this in mind, and consciously protect the individual by anonymising data or de-identifying people at the edge, you take a big step towards making cities not only smart but also private.

In fact, AI can actually be used to make public spaces even more private than they are now. In other words, AI itself with the right constraints can become a privacy technology. Consider a camera used for parking or traffic enforcement. Traditionally, it captures everything: the relevant vehicle or incident, but potentially also pedestrians, children and other people who have nothing whatsoever to do with what is being enforced.

Cities should require and widely publicise policies that prefer sensor data fusion or physical AI systems that mask people’s faces, including the faces of children who may accidentally appear in imagery. The system can retain the information necessary to understand the event while deliberately obscuring information it doesn’t need.

We have heard a lot about technology ‘escaping’ and running amok. Some have said that we risk letting AI become a civilisational risk. But this apocalyptic thinking takes humans not only off the hook, but out of the loop, and we’re not — not by a long shot.

Technology itself still doesn’t make any decisions, at least not for long. We do. And sensors used appropriately can deliver AI for good. I was reminded of how tangible this can be when SenSen began working on a problem I had badly underestimated: urban noise. Until we became involved in developing AI-powered noise enforcement technology with the EPA, I did not appreciate how profoundly excessive vehicle noise can affect people’s lives.

Talk to people who live with it and this stops being a discussion about decibels. It is someone being woken at two in the morning by a modified exhaust. It is a family repeatedly disturbed by motorbikes roaring through a residential street. It is the intimidation of burnouts, engine-revving and deliberately antisocial driving in neighbourhoods where conventional enforcement simply cannot be present all the time.

Another common problem for cities and suburban neighbourhoods is heavy vehicles running their streets. While you might see cameras being torn down around the world in anti-surveillance protests, residents on streets affected by heavy-vehicle rat running beg for the cameras back if they are ever removed.

This is why I don’t think the right measure of a smart city is the number of cameras it has, the amount of data it collects or even the sophistication of its AI.

A truly smart city should be judged by what changes for the person living in it. Did I spend less time sitting in traffic? Was my child safer crossing the road? Did an ambulance reach somebody faster? Was the air cleaner? Did I get a decent night’s sleep? And critically, did the technology achieve those things without unnecessarily identifying, tracking or remembering me? The goal is not more AI. It is less congestion, fewer accidents, quieter nights and cleaner streets while maintaining greater privacy. Technology done right in this context not only makes a city smarter, it makes it more human. Now isn’t the time to rip out cameras and sensors; it’s the time to demand and create how they operate in the most human and humane way possible.

Top image credit: iStock.com/Andrey Suslov