AI and the second generation of smart cities

Posted: November 15, 2024

In January 2010, Sam Palmisano, CEO of IBM, came to the venerable British think-tank Chatham House and delivered a speech entitled “Welcome to the Decade of Smart.” In it, he outlined the now-familiar vision of a connected world: IoT-enabled sensors were generating vast quantities of data and powerful computers were turning that data into knowledge. “With this knowledge,” said Palmisano, “we can reduce cost and waste, improve efficiency and productivity, and raise the quality of everything from our products, to our companies, to our cities.” ^[1]

The breadth of Palmisano’s vision was at once inspiring and common-sensical. In 2010, the hardware-software combination he was describing already had a proven track record in the corporate sphere, and cities—just like companies and factories—rely on complicated and interlocking systems, processes and machinery, all operating at various levels of (in)efficiency.

“The smart city agenda aimed to impose a measure of rationality on twenty-first-century urbanism,” Canadian journalist John Lorinc wrote in his 2022 book Dream States. “Certainly, to those charged with governing and administering cities, the technocratic promise of the smart city has been highly appealing. After all, what city wouldn't want to be ‘smart?’” ^[2]

After an initial spike in public enthusiasm, however, the extent to which a city differs from a company became impossible to ignore, and the seemingly innocuous vision of the “smart city” was sucked into a morass of local politics, privacy concerns, and tech-skepticism.

No project better encapsulated the PR travails of smart cities than Google’s infamous efforts to redevelop a stretch of Toronto’s waterfront “from the internet up.” Sidewalk Labs went public with its proposals in late 2017, but in 2020—after two-and-a-half years of controversy and backlash, and in the midst of new economic uncertainty from the COVID-19 pandemic—the project was abandoned and Sidewalk Labs was absorbed by Google.

Meanwhile, another smart city debacle was unfolding in San Diego. In 2017, the city installed cameras and sensors on thousands of streetlights with the stated purpose of saving energy and gathering data on parking and bike lanes. When it emerged that the police department was using footage captured by the cameras in criminal investigations, San Diego citizens were outraged by what looked a lot like unchecked, secretive surveillance. All 3,200 cameras were switched off in 2020.

Palmisano’s 2010 prediction about a “decade of smart” ended up being uncannily accurate, although not in the way he foresaw. By 2020, as cities contended with a global pandemic and an increasingly privacy-conscious citizenry, it was clear that smart city technology was not the friction-free fix-all it had seemed ten years before.

People as assets; data as infrastructure

At the heart of both the promise and pitfalls of the smart city is a vexing question: where does infrastructure end and where do people begin?

Take a classic target of smart city technology: mobility. A “smart road” might be monitoring things like traffic flow, average speeds and the locations of collisions, and the data generated might be used to dynamically adjust speed limits, road lighting or traffic light schedules. Roads, lights, signage—these are quintessential pieces of infrastructure. But the real objects of attention for a smart road are the vehicles using it. Are those vehicles part of city infrastructure? Or are they merely proxies for the people driving them?

In this case—and in many others—drawing a clear line between a city’s infrastructure and its people is simply not possible. From there it follows that, as one commentator put it, “smart cities are surveilled cities.” Once this basic fact is admitted, governance questions around the data being collected immediately arise: Who owns it? Who has access to it? What decisions is it informing? Who is responsible for these decisions?

In Toronto, the answers to these questions were so unconvincing that Sidewalk Labs’ own privacy consultant resigned. But in Singapore the debate around data was more productive.

“When we first started the journey, I think many people were very enamored [with] just doing a lot of apps,” explained Cheong Koon Hean, Chair of the Centre for Liveable Cities in Singapore, in a 2021 talk with the Institute for Management Development. One of the main challenges Singapore faced, according to Cheong, was “winning hearts”: “How do we give people assurance that their data will be protected? We had to put in place what I call these ‘critical enablers,’ like policies and standards and regulations.”

Far from hampering the adoption of smart city technology, such civil groundwork paved the way for the completion, in 2023, of a cutting-edge digital twin of the entire city-state.

Singapore’s virtual model of itself—and all the use cases such a model might have—highlights the extent to which the data collected in smart cities has the potential to define the parameters of urban life as profoundly as water and power. Indeed, scholars have characterized data as “a key urban flow that underpins a new type of urban utility,” which they term “data-as-infrastructure.” ^[3]

Community and AI

One might have expected the release of ChatGPT in November 2022 to spark a relitigation of the “decade of smart.” Generative AI, after all, seemed tailor-made to reconvene all the original plaintiffs and defendants: here, again, were widespread data collection, under-theorized technology, promises of transformative efficiency and the potential for unaccountable decision-making.

Instead, many cities seem to have learnt valuable lessons from the past decade. In 2021, the Mayor of London’s Office published An Emerging Technology Charter that provides “a set of practical and ethical guidelines for the trialing and deployment of new data-enabled technology.” In 2023, New York City published its AI Action Plan. Both documents recognize the need to actively engage the public and for clear regulatory frameworks.

Speaking at a recent industry conference, John Paul Farmer, the CEO of broadband company WeLink Cities, distilled the consensus around AI-powered city technology: “Sometimes maximizing privacy actually means less equity or less security. Having those hard conversations is how you actually come to understand what the priorities are for your community and make sure that you’re using AI responsibly in the context of what that community wants.”

The tenor of Farmer’s remarks is echoed in the new community-centric innovation initiative launched by trade publication Smart Cities World, which urges technology providers: “partner, don’t dictate” and “prioritize privacy and security by design.”

Though a far cry from the sweeping, utopian vision presented by the IBM CEO back in 2010, today’s industry-wide interest in community and privacy bodes well for the next era of smart cities. After an adolescent decade of rapid growth and frequent embarrassment, the smart city industry is carrying itself with the self-awareness and humility of a full-grown adult.

References

^{[1] Transcript: Welcome to the Decade of Smart
[2] Dream States: Smart Cities, Technology and the Pursuit of Urban Utopias, John Lorinc, p17
[3] Urban Operating Systems, Andrés Luque-Ayala and Simon Marvin, p6.}