Solving Big Problems With Big Data
Angela Weiss / AFP / Getty Images

Our brains weren’t built for city scale. AI can help.

New York City is noisy, dense and unapologetically ambitious. It was built by waves of immigrants, strivers, startups and merchants who turned a crowded island into the world’s economic engine. 

New York’s density necessitated urban innovation. The modern grid, the subway system and the skyscraper were all essential to make room for our ever-growing population and ambition. 

Today, cities are among the world’s most dynamic, livable and equitable places because they bring so many diverse lives into shared space. But this also means that they are confronting the next generation of wicked problems, from housing affordability to coastal flooding. 

On our own, we might struggle mightily to address these and other dense, interconnected challenges. Human biology designed our brains for three-dimensional environments. We struggle in four dimensions when considering climate change and its long time horizon, for example.  

Yet with help, rich datasets can help guide us toward answers. Every street corner links vehicle flow, transit delays, shadow footprints, storm runoff, retail sales, mass timber supply chains and power grid loads. AI excels at digesting variables like these, analyzing what they mean, tracking fluctuations over time and surfacing patterns and connections, much as we have seen with AlphaFold’s ability to decode complex biological systems in the medical sphere.

First, when defining artificial intelligence, it’s important to recognize the vast array of technologies beyond the generative AI chatbots we may be most used to. In addition to generative AI helping us summarize text or write computer code, these large models also help us with computer vision, protein folding, projection and many more areas, but their real skill is in semantic context-mapping of natural language — from transit telemetry to policy text — in a single, coherent network. This high-dimensional insight allows us to map hidden connections across complex urban systems, and gives us the computational power to simulate potential solutions. As a result, AI offers cities an unprecedented opportunity to expand their capacity and lead the next wave of urban problem-solving. 

My career path has given me a unique window onto the opportunities that AI presents for cities. For 20 years, I have worked at the intersection of technology and social impact across health, crisis response, education and urban design. Today I serve as head of Data Commons by Google, building an open statistical knowledge graph to ground AI models in verified public data. 

Before this role, I served as president of urban products at Sidewalk Labs, where I led our technical teams to build the tools that would enable affordable, equitable, climate-positive urban neighborhoods. We were incredibly ambitious, and while we didn’t build the city of the future, we pushed the boundaries and helped forge a new blueprint for future cities. Some of those innovations have already reshaped the modern urban landscape, from generative neighborhood design to smart building-energy controls, while other foundational ideas are only now beginning to realize their full potential.

That work builds on the vision Dan Doctoroff championed in New York City and Toronto, guided by Sidewalk Labs’ core belief that “by combining people-centered urban design with cutting-edge technology, we can achieve new standards of sustainability, affordability, mobility, and economic opportunity.”

One of the biggest challenges cities face today is data fragmentation, which keeps cities blind to their own complexity. 

But when you look at municipal challenges through an engineering lens, the areas of health, transit, housing and climate resilience are all nodes in a single network. Small changes can create massive downstream ripple effects. Changing a building envelope or narrowing a road by two feet shifts vehicle speeds, bus turnaround times, bike safety, solar shadow access, storm runoff and sidewalk retail foot traffic. 

Human minds alone cannot trace those cascading multidimensional effects manually. But human minds in collaboration with AI can. 

Fundamentally, AI is a tool. Like all tools, it exists to extend and enhance our human capabilities. When faced with multidimensional urban challenges like these, AI gives us the insight to uncover hidden connections, and the computational muscle to virtually simulate thousands of iterations before pouring immovable concrete.

We’re already seeing this transformation today in spatial modeling, as demonstrated by Flood Forecasting, which uses applied AI, high-resolution satellite imagery, water trends and more to warn 2 billion people in over 150 countries in areas at risk for significant flood events. In an urban context, you can now design new buildings or solar footprints, while projecting everything from project cost to solar generation, with a few clicks in Google Earth. Historically, advanced modeling was locked behind expensive, proprietary software and multimillion-dollar budgets, putting high-level analysis out of reach for most communities. Urban planning is now accessible to small city departments and local community boards. Planners can run generative spatial scenarios in minutes to evaluate density, tree canopy cover and solar access. 

AI coding agents go a step further, allowing anyone to build their own spatial models and run complex simulations without even needing to know how to code. Within a generation, as computer vision and AI skills become everyday utilities, every city planner will gain superpowers to identify, diagnose, predict and fix numerous urban issues. 

Take the curb as an example: At Sidewalk Labs, we developed scheduled flex zones that could be orchestrated dynamically by a multimodal AI system, adjusting traffic flow and parking to street activity in real time. All the technology pieces for such a system exist today, but bringing them to life requires deep cooperation and a shared vision among city leaders, technology partners and the public.

Of course, if AI is a superpower, then we also need to bear in mind the lesson first popularized by one of the most famous New Yorkers, Spider-Man: With great power comes great responsibility. For urban AI, this means new privacy regulations and controls, because public trust breaks down when AI technology turns into public surveillance. 

We confronted this challenge directly in Toronto. To ensure information was handled in a privacy-sensitive way, we proposed an independent Urban Data Trust to serve as a regulatory body overseeing all data collected in the public realm. We took a strict “privacy by design” approach, proposing that anyone wanting to deploy a sensor had to fill out a Responsible Data Use Assessment outlining exactly how they would protect individuals. The goal was to treat privacy like software. Regulators could run automated unit tests against the network to verify a device was stripping out personal identifiers before anything ever reached a server.

Making that technical oversight possible requires systems built entirely on open standards and open application programming interfaces (APIs) so independent groups can audit how they communicate. These same open systems could also encourage cities to procure smaller, cheaper components to ensure they can stay on the innovation cycle (and not fall behind), and speed up the request-for-proposals process by using software concepts like unit tests to help approve such devices. 

Transparency also has to work on the street. People walking through a neighborhood deserve to know what is happening in the physical space around them. We helped develop the open-source Digital Trust for Places & Routines (DTPR) standard to solve this visibility problem, and it continues to live on as an independently stewarded open standard for the industry. 

Through deliberate and thoughtful governance, cities can collect operational data without sacrificing privacy. For example, cities can require sensors to remove faces, license plates and device identifiers before data ever reach a server. Public spaces should remain free of facial recognition. Cities should require open and visible signage (as proposed by DTPR) wherever sensor technology is in use. All this data should be published through an open data commons, so local developers and residents can build their own solutions, because data collected in the public realm must be treated as a public asset. 

Grounding AI in open data with responsible privacy protections isn’t a constraint. Rather, it will create the foundation for future progress. When we give our city builders, government officials and residents access to this bold new technology, we can solve today’s intractable wicked problems. New York City can continue to be the world’s most dynamic, livable and equitable city while setting a global standard for responsible urban AI.


Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to Vital City.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.