Editor’s Note: In Part 3 of the Free Cities Center series, Serlet looks at an AI program that offers some benefits, but raises much more serious questions.
Hangzhou’s City Brain
In 2020, the city of Hangzhou in China announced that it had developed a “City Brain.”
Hangzhou is an ancient city with a population of 13 million. When China developed, Hangzhou became a center for manufacturing. In recent years, the city has attempted to pivot by focusing on livability and tourism rather than manufacturing.
Hangzhou’s City Brain is an artificial intelligence that gathers billions of data-points. The data comes from municipal records, tax records, census, police reports, IoT sensors, traffic cameras, toll stations and many other sources. The data is centralized and can be queried to answer any specific questions.
The most important aspect of City Brain isn’t that it collects data per se. Instead, it’s that Hangzhou is future-proofing itself for future AI developments by pre-collecting and pre-organizing training data. As AI technology improves, Hangzhou will be able to use the training data within City Brain to immediately implement the new technologies.
The project was mostly financed by the city of Hangzhou’s municipal budget for infrastructure development with some limited grants from the national government. Hangzhou opted to work with a single vendor to develop the entire project: Alibaba.
The Hangzhou City Brain has already had some success stories. The first use of the City Brain was to decide how officials would rework the city’s traffic network. In 2020, Hangzhou was the second-most congested city in China after Beijing. Officials used data from the City Brain to determine which traffic lights to add, where to add safety equipment, and how to concentrate the efforts of traffic police. Traffic has significantly decreased and Hangzhou is now the 34th most congested city in China. Accidents also significantly decreased.
Prior to the development of the City Brain, citizens had to spend significant amounts of time in government offices to accomplish basic tasks like registering new home purchases or applying for permits. The government created online portals for all services, and used the City Brain to determine whether or not the portals were successful. Now 97% of all government services can be handled online.
During COVID-19, the government also used the City Brain to organize quarantines, rewrite the health code and plan response policies. The long-term goal for the City Brain is to collect data for future use as AI models become more sophisticated. Now the City Brain is being used to train language models and chat-bots. In the future, officials hope it can be used to train artificial general intelligence. The project has been so successful that 23 other cities across Asia are now implementing their own “city brains.”
If the Chicago Department of Public Health’s use of algorithms to determine which restaurants to inspect is troubling for liberty, and California’s mobile driver’s licenses are troubling for privacy then that raises many obvious questions about the concept of a city brain.
City brains are the future. Language models already need large amounts of training data. Pre-storing municipal data for use in future AI developments is an obvious solution to that problem. Many AI experts believe that over the next 15 years we will develop Artificial General Intelligence (AGI).
AGI is a technology that can make autonomous decisions. For example, in the future AGI could be asked to reduce crime in an open-ended fashion. AGI would then be able to create a plan to do so and execute that plan. Of course, AGI is still theoretical and a matter of science fiction. Cities that have pre-centralized all data into city brains will be the first to benefit from AGI.
Should a technology like AGI come to fruition, Hangzhou’s City Brain could produce truly miraculous results. And those results could very well be extremely creepy.
Conclusion
The era of human decision making is coming to an end.
In the future, more and more municipal services will be fully automated. Automating municipal services is good. But all progress comes with risks. The main risk is that automation will destroy citizens’ ability to make decisions for their own lives. Instead, decisions will be made for citizens by machines – and citizens will be forced to comply using technology such as digital identity.
There is also a possibility that these concerns will prove unfounded. One survey of 100 cities found that while 96% of mayors wanted to use generative AI, only 2% of cities were actually using generative AI. There could easily be a future where there is a lot of talk, a lot of hype, a lot of small scale prototypes and very little large implementation.
A future without automation is not a good future. Although it comes with less existential risks, it also comes with bureaucratic sloth and municipal inefficiency.
AI ethicists are attempting to create rules and algorithms to prevent these problems. The problem is that morality is not an algorithm. Freedom is an abstract principle which is hard to define. We need to make sure that municipal institutions place a higher value on subjective human moral judgement calls over automated decisions.
Thibault Serlet is a partner at Key State Capital, a Venture Capital consortium that invests in digital identity technology. He previously served as the president of the Adrianople Group, a business intelligence firm which helped investors finance the creation of new Special Economic Zones. He led the creation of several large scale datasets including the Web of Trust, a global database of decentralized digital identity projects, Open Zone Map, the world’s first global map of free zones, and the Charter Cities Institute’s New Cities Map.
