Generative AI is quickly gaining traction across industries, including urban planning, where the potential for automation and data processing promises to transform how cities are built. In Amsterdam, where the urban landscape constantly evolves, researchers at AMS Institute—Lydia Giokari of Urban Energy and Fabian Geiser and Thijs Turel of Responsible Sensing Lab—have investigated how AI can be applied to city planning, tying into the current curiosity and caution around using AI in this field.
Their goal is to understand where generative AI could enhance urban planning processes while considering potential drawbacks, such as loss of local identity, risks to data privacy, bias, and the hefty environmental impact of using AI technologies.
“My dream would be that Generative AI could be used widely as a tool to both empower and increase direct citizens’ participation in the (re)development of the physical environment of cities. I would love to see Amsterdammers using their phones to scan locations around the city in real-time in the future, adding their wishes, dreams, and preferences and translating them into design solutions for their streets, neighborhoods, and public spaces.”
Lydia Giokari
Program Developer Urban Energy
Generative AI is an Artificial Intelligence method that includes tools that can create content based on patterns or large datasets. Generative AI for urban planning refers to the use of advanced AI models, such as machine learning algorithms, to automatically generate design options, plans or solutions for urban development — ChatGPT or Leonardo AI, for example.
“Generative AI is all over the place, so we really need to think through how this affects our everyday job of designing and planning the city. There are huge opportunities to expand our creativity and productivity, but we also need to be cautious about potential pitfalls and even dangers. Designing and planning is a human job, AI is a tool that has the potential to give great power to those humans but it comes with great responsibilities as well.”
Jan Duffhues, Teamleider Beeld & Data, Ruimte & Duurzaamheid bij Gemeente Amsterdam
How could generative AI tools be used?
Generative AI tools, widely used globally, are now being tested in Amsterdam's urban planning processes. AMS Institute attempted to link the possible generative AI applications to the urban development process stages known as the Plaberum: a structured, phased approach to the urban development of an area that includes four phases: Exploration, Feasibility, Design, and Execution. Each step is rounded off with an administrative decision directing the next step.
The research highlights that Generative AI tools and applications can enhance the Plaberum process in different ways: speeding up workflows, providing rapid scenario modeling, visualizations, and predictive analyses, for example, thereby helping planners and stakeholders visualize future developments and make data-informed decisions. By simulating different design outcomes and forecasting impacts while involving local stakeholders, these tools can support more inclusive and dynamic urban planning, aligning with Amsterdam's sustainable and participatory urban growth goals.
“The research shows that AI can help us improve our current interaction and participation with citizens. For example, by translating their needs into design, filling an AI training database with personal photos of their neighborhood, or acting as a participation entity. We are exited to explore the opportunities”
Debbie Dekkers, Innovatiemanager & i-team director Digital Urban Planning bij Gemeente Amsterdam
In collaboration with the Municipality of Amsterdam's departments of Planning & Sustainability and of Innovation, AMS Institute presented possible application opportunities already in the pilot stages. These range from planning support to visualizing data for municipal purposes (see below).
Speeding up workflows, but threatening livelihoods and generating “cookie-cutter” cities?
As Amsterdam aims to incorporate AI into its urban processes, questions arise about whether it could replace human expertise and what that means for municipal employees. While AI might help with data management and decision-making, there is a delicate balance between automation and maintaining the human touch needed for unique city planning.
Building on this, the research also cautions that relying heavily on AI could lead to generic urban designs that lack local character. Cities are inherently messy, with diverse histories, people, and cultures reflected in their infrastructure: generative algorithms might smooth over this uniqueness. Careful management is required to avoid creating "cookie-cutter" cities and give space to cities' and their citizens' inherent 'messiness'. Privacy concerns, biases, and data security must also be considered, especially in an urban context where large amounts of data are handled.
AI's high energy consumption and environmental impact, especially its use of materials, water, and energy, are also rightly being scrutinized. This comes down to the fact that AI models require GPU-intensive generative AI systems, which demand substantial electricity use, resulting in a high carbon footprint. Simultaneously, the demand for continuous cooling, for example, at data centers, further escalates the energy use footprint, as well as water consumption. The vast amounts of water needed to maintain optimal operating temperatures in data centers is increasingly impacting local water resources, especially in regions already facing water scarcity. Furthermore, the hardware used to run large-scale AI systems rely on scarce and limited material resources, whose extraction and refinement can have negative environmental and social impacts.
Expanding AI is a concern at a time when six of the nine planetary boundaries that measure environmental health across land, sea, and air have been crossed. Resource demands must be critically considered when deciding whether to use these technologies over others and focus on sustainable practices in AI development where possible.
“The energy, water, and raw materials required for AI processing raise serious concerns. Estimates show that AI-related energy consumption could be up to 20 times more than traditional forms of processing, and as AI chips continue to evolve, we could see millions of tons of e-waste by 2030. My hope is that in the future, we will have ways to easily compare the benefits and drawbacks of generative AI, enabling us to make informed decisions and maximise its usefulness without sacrificing our environment”
Fabian Geiser
Project Manager Responsible Sensing Lab
Moving forward, help us investigate the delicate balance needed?
This preliminary research shows that while AI can help the Municipality of Amsterdam process complex data and manage urban development, careful consideration is needed to maintain the city's unique identity and to balance the work with environmental goals.
We’re keen to continue exploring this field: Researchers, policymakers, and industry professionals are encouraged to contact us, explore further research, and submit proposals to help shape the future of AI in urban planning!
