2019 Industrial AI Sketch: Urban Edition
-
In 2020, urban AI initiatives will undoubtedly make some progress. Objectively speaking, this progress will primarily stem from cloud services penetrating deeper into urban and government markets, offering more specialized and targeted AI solutions.
When it comes to the changes AI has brought to cities, I might have some authority to speak.
I live in Tianjin Eco-City, which is essentially an island. Every day, I have to cross a bridge to get out. Naturally, everyone tends to leave at similar times, so by 8 AM, the bridge is predictably jammed. What used to be a 30-minute delay has become a reflex. But after this summer, the congestion has significantly decreased, and often there’s hardly any traffic at all.
Later, I discovered that the traffic lights at major intersections in Eco-City had been taken over by AI. In the mornings, traffic is concentrated on the main roads, while side streets are nearly empty. Smart cameras adjust traffic light timings based on real-time vehicle flow, allowing me to inexplicably sleep an extra half hour.
I travel frequently to Shenzhen, and anyone who’s been there this year would surely notice the changes at Bao’an Airport. The obvious change is facial recognition for check-in and security—frequent flyers can breeze through fast lanes, an almost surreal convenience. The less visible change is AI algorithms optimizing gate assignments, drastically improving boarding bridge usage. Flying to Shenzhen now rarely involves the dreaded shuttle bus, though flying back to Tianjin is another story… All this is thanks to AI.
If you live in a first- or second-tier city and pay attention to infrastructure and municipal news, you’ve likely noticed how AI has integrated into urban life in 2019. At the very least, facial recognition for high-speed rail boarding or convenience store payments is hard to miss.
In the past year, AI has genuinely transformed urban living and reshaped the already massive 'smart city' market.
To document the changes brought by urban AI, we might need to start by clarifying a concept.
By 2017, over 500 Chinese cities and regions had explicitly proposed smart city initiatives, with more than 3,000 projects underway. But think about it—2017 was just the dawn of the AI market, and AI’s integration with cities was mostly conceptual.
That’s right—those 3,000+ projects had little to do with AI technology.
So how did cities become 'smart' without AI?
This traces back to 2008, when IBM introduced its 'Smarter Planet' initiative. While 2008 doesn’t seem that long ago, we’re now in the 2020s—it feels like recalling the 1980s in the early 2000s, ancient history.
The so-called Smarter Planet encompassed various technologies but essentially involved collecting data on urban traffic, water systems, buildings, etc., and creating data visualization systems. IBM’s main goal was to sell IT infrastructure to governments after the corporate IT market became saturated.
This, along with the subsequent rise of smart city projects, should more accurately be called 'information cities.' They merely aggregated urban data for managers, making them 'smarter'—not the cities themselves.
Of course, this approach had value, but it didn’t make cities inherently smarter. Later, smart city projects became even more bewildering—any government-led project was labeled as such, even creating a public account for a department.
It wasn’t until AI arrived that things changed. Whether it’s facial recognition for train stations or smart cameras identifying fugitives at concerts, even the simplest AI+city projects at least replaced manual labor with machines. Only then could cities rely on machines for intelligence.
However, after over a decade of smart city development, rebranding isn’t practical. Moreover, smart city projects often serve as the foundation for AI integration, as data collection and organization are prerequisites for AI to function.
This leaves AI in an awkward position within smart city frameworks: On one hand, governments and citizens are so accustomed to smart city projects that they hardly notice AI-driven changes. On the other, AI+city is inherently complex—a collection of countless disparate projects. For example, environmental monitoring and urban security rely on machine vision, while rail management and airport automation are data-driven, and utilities like water and power are essentially industrial AI projects.
Thus, terms like 'smart city,' 'AI+city,' and 'urban intelligence' often devolve into semantic games, with many companies exploiting the confusion. This ambiguity makes it hard for industries and investors to gauge the value and potential of urban AI, leaving the field dominated by established players rather than startups.
If you can’t quite distinguish between 'smart city,' 'intelligent city,' 'city brain,' and 'AI city,' you’ve identified the core issue: Even the clients commissioning these projects likely can’t tell them apart.
Today, the landscape includes non-AI smart city projects, AI-enhanced urban informatization, and AI-led urban initiatives—all operating without unified standards. Still, 2019 has brought many surprises in AI-city integration.
Earlier, AI’s urban applications were largely limited to traffic. So-called 'city brains' often turned out to be AI-controlled traffic lights on a few streets. While valuable, this barely scratches the surface of urban intelligence.
By late 2018, comprehensive industrial AI solutions emerged. In 2019, AI began appearing across Chinese cities. Though full-scale urban AI remains distant, we might call this phase the 'AI+city niche' boom.
For instance, weather bureaus in Beijing, Chongqing, and elsewhere deployed AI-meteorology systems in 2019, using deep learning to improve short-term forecasts. This year’s rapid disaster warnings in southwestern China were powered by AI.
Similarly, smart cameras now routinely identify violations and pollution. Suzhou’s water authority uses machine vision to monitor waterway pollution, enhancing environmental responses.
In Ningbo, smart city management systems use AI to detect illegal parking, debris堆放, and unauthorized vending, pinpointing violations instantly. Fire departments combine AI cameras with alarms to predict fires and optimize rescues.
Many urban systems—heating, power, water—are essentially industrial AI challenges. These now employ predictive models and AI quality control to boost efficiency and safety.
Beyond traffic lights, AI now powers street monitoring, smart bus stops, and highways. Autonomous driving and vehicle-infrastructure coordination are also being tested.
Ultimately, cities are vast networks of functions, infrastructure, and labor. Though urban AI lacks standards, every detail represents a market opportunity. Making each corner of a city 'AI-enabled' is a multi-year endeavor.
In 2020, urban AI is poised to make certain advancements. Objectively speaking, this progress will mainly stem from cloud services penetrating deeper into urban and government markets, providing more specialized and targeted AI solutions.
Additionally, the IoT market is approaching industrial maturity and standardization, which will offer hardware support for numerous urban intelligence needs. This also signifies new development opportunities for urban AI.
In other words, urban AI in 2020 is more likely to experience quantitative growth rather than a qualitative leap. However, we can still observe that the overall approach to smart city construction is evolving, with comprehensive urban intelligence becoming a new focal point in discussions.
Earlier, during the development of 'government cloud' and mobile government channels, the industry quickly identified a problem: various government departments independently built their own systems, creating isolated private cloud infrastructures. This essentially led to siloed systems where data was not shared, systems were incompatible, and services operated independently. This not only increased the learning and usage burden for the public but also hindered the long-term development and unified planning of government cloud services.
In the development of urban AI systems, siloed structures may soon re-emerge. Water management AI and logistics AI might remain incompatible, traffic AI and aviation AI could stay disconnected, and even different districts, institutions, and schools within the same city might develop their own AI systems.
The issue with this model is that AI development relies heavily on the continuous input and reuse of data, and urban data is often interconnected. For example,人流 data from shopping malls, schools, and stadiums directly impacts traffic and rail transit data, while smart energy-saving data from residential areas and factories can influence intelligent energy management for the entire city.
Based on this prediction, comprehensive urban intelligence should be the future direction. Only by enabling data sharing and collaborative intelligent decision-making across departments can cities achieve a true 'urban brain.'
Currently, there are three key opportunities for urban AI to evolve toward comprehensive intelligence:
Ultimately, urban AI is part of the broader urban network, cloud computing, and big data ecosystem, relying on decisions that promote overall urban intelligence. Urban intelligence is a top-down project, and sound decision-making is its greatest driver.
At the 2019 Smart City Expo World Congress in Barcelona, Yingtan, Jiangxi, won the Global Smart City Digital Transformation Award. This lesser-known city achieved international recognition by embracing IoT development and deeply integrating urban infrastructure with IoT technology.
This demonstrates that urban AI and technological integration are not exclusive to economically advanced cities. Instead, smaller cities may find it easier to 'start fresh,' gaining new development opportunities.
Recently, the concept of 'China's dividend' has gained traction. While it's unclear how many industries will benefit, in the field of urban AI, this dividend is undeniably real.