How Is AI Transforming Urban Design and Planning?
Listen to the full episode
Meet Damiano!
Can you tell us briefly about your AI platform?
The platform is about using generative AI to codesign the future of cities. It’s not just a platform in the sense of a technological platform, but it's also a method. Participatory AI is something new and the overall idea is that instead of having a design process that involves different stakeholders in different parts of that process, you have them all together. The job of AI is facilitating the codesign phase. That usually means a project, a renewal project, an urban design project or even a new street. And our main users are the cities, ministries and architecture offices.
How long have you been involved in AI?
I cofounded a company called SPIN Unit over 12 years ago, helping cities and ministries do research. At the same time, I met Liza Gazeeva and Luca Stornaiuolo, cofounders of the AI company Toretei. They were already training their own generative AI models. We started working with cities that wanted to have new tools design alongside citizens and other stakeholders. Basically, five years ago, the current project started.
So how can AI help with the design process?
This dialogue between planners and stakeholders is the most important but often we don't even have that conversation. It's very difficult for planners to have that kind of conversation so quite often people work in ‘black boxes’. One person is taking care of one design aspect, somebody else is taking care of the participatory planning, another person is taking care of other elements and then they try to bring them together. The problem is that these groups have their own methods.
AI can help participants ideate and evaluate what the future could be like. If you think it might be great to try a small urban pocket-park here, but you have no idea how it would look, that's something usually would take months for architecture offices or weeks to show. But with AI you can have an image in the same instant and you can decide right away if it works or not.
Can AI help involve the public in urban planning?
What AI can do, at least in our case, is help people image the future, which is the most important thing when we are talking about cities, but it’s also one of the biggest hurdles. If, for example, you ask somebody what ‘works’ or what ‘doesn't work’ in their city, it is very easy to answer, because they have a lot of lived experience. But if you ask somebody, how would you like this square in front of us to be in the future? That's a very difficult question because the possibilities are almost infinite. What AI does is use the universal language of images, which can add a vocabulary that everybody can use. They can basically talk about different areas and speak to each other and visualise the results. That is the main advantage of AI. It allows us to change our mind. Quite often you might want something, but then once you see it realized, you might not like it anymore.
How can the public be involved in the planning process?
Take, for example, mobile phones. A member of the public can take a photo of an urban place. They can then select with their fingers the area they want to transform, and then let the AI help them to come up with ideas. You can just go around the city anywhere, take a photograph and express your ideas.
I can describe the process we've done for the for the Helsinki summer streets. There was a workshop in January, which was a representative group of citizens. The idea was that the participants, citizens plus designers, plus city representatives, would come up with what they wanted. They took their ideas to actual photographs which they can then evaluate. If they wanted a bike lane, how would it look? Then they could take a decision.
It can also allow us to see any problems or issues. One thing that people wanted were bike lanes on one of the main central streets. However, when AI ‘built’ an image of a bike lane on the street, which showed that the bike lane is going to basically destroy the 19th century cobblestones, people then said they didn’t want the bike lane anymore. They wanted to find a new solution. When you understand the real impact of a change, you might want to change your mind and improve it in a different way.
And this gives stakeholders agency. When they took the citizens and the participants back to see the constructed area, they basically said that, ‘Yeah, this is a bit different of course, but we see our ideas in it’.
Does AI make the urban planning process more efficient?
Take the example of what happens when you have an online campaign. You could collect 50,000 ideas; however, many are contradictory. We can use the AI to show the most common, or even the most diverging options as well. You try to understand which ideas people agreed on, and also the outliers as well.
What’s the main motivation for ‘speeding up’ urban planning?
The reason why we want to speed up the process is because we have to. There are about 26 new EU directives that are basically informing us how should we regenerate our cities in order to be resilient, both in terms of climate resilience but also economic resilience. That's where the where the rush comes from. It's not that we want to speed up just because we want cities to grow; instead, we want to retrofit them according to international frameworks. There are a huge number of lists of both directives and guidelines from the EU. We are in a rush to retrofit our cities. Therefore, we need to be able to include not just multi-stakeholders but also understand how different species live in cities.
I would say that AI makes the process faster by using collective intelligence of our cities to reimagine them. The point is to change the process. If you can collect multiple intelligences when you actually are in the design phase, it is most likely that you will have innovation. That’s the most important thing. Innovation usually comes when people with different ideas come up with a third idea. What AI does is make that conversation easier and faster. You can do this process without AI if you want to, it's just going to be so much slower.
Additionally, you can increase the quality of the work that you do. What we're seeing is that the engineers and architects have to be compliant to the regulations of cities. With AI, they can actually focus so much more of their time on the quality of their own process, rather than exactly producing many different scenarios and trying to please everyone in different stages.
Can AI help evaluate how a building complies with building regulations?
This is one of the things we're working with a lot. We are training the generative AI model based on either policy, guidelines, design guidelines, or even desired regulations. This is one of the most important processes because somebody might produce an image, but they might only judge it in the context of a conversation or for its aesthetic properties. But aesthetics is just one element, regulations and even technological possibilities are just as important. We can train AI according to specific criteria or specific local regulation.
How do you see the future of urban design and AI?
We are involved in many projects with policymakers. One of the most interesting is with the United Nations Development Program. We’re introducing UrbanistAI to about 250 municipalities from Eastern Europe. It's called the Mayors for Economic Growth. We're actually helping the mayors and the mayor's offices understand how to utilize AI to codesign policies, and also to understand policies.
Let's say you're a mayor of a city and you get an EU framework or an EU directive. You need to take the document, interpret it, and try to understand what it means for your city. What does it mean for the policy direction? You can ask AI to help you to do that.
One example is in Estonia. There is an exhibition at the National Museum of Natural History that allows basically participants or kids, whoever comes to the museum, to understand the EU policies on nature-based solutions. They go into a museum area where they can choose a photograph from the city of Tallinn. And then they can apply a policy. What if we increase the biodiversity of this trade? What if we apply nature-based solutions to this? What if we build the BIOS wave here? So, it's a way for them to really understand what this policy means without even knowing that there is a policy.
Will AI produce a uniformity? Will urban planning become standard across the world?
You can actually use synthetic material as a training model. Take an example from Dubai. We've been working with the Prime Minister’s Office there to create the generative AI model for Dubai. But often, if you use a generative model, it's going to produce a new city you want, or a new street or new park but it will create an image that looks like Dubai. However, in planning you often want to design something that doesn't look the way it always looks. Sometimes they want to say ‘I want to produce something that doesn't look like Dubai’.
Then you start to produce synthetic training sets. Which means that planners, who might have some ideas and politicians who might have different ideas, give their perspectives of the future. Then their ideas become synthetic images that are used to train the models. Basically, AI is going to learn from visual features. When you when you use UrbanistAI, you can decide if you want the model to be conservative or creative, futuristic, or something else. This kind of synthetic training is the most interesting part.
Is AI going to replace urban planners?
Well, I would say that a lot of planners see this as a way to have faster processes. They don't see it yet as ‘it's going to take my job’, they see it more like, ‘if the planning process is faster, so be it’. Some people are, of course, concerned, mostly designers.
However, AI is only going to take our jobs if we design a process for it. If you want, you can redesign the process for AI to basically replace the planners. You just collect ideas then these ideas are turned into prompts. The prompts are turned into drawings, then you might connect it to another AI that creates a model out of it. The question is, is this a process we want? I think many people don't want that.
That is the kind of conversation we should be having now. What kind of processes do we want to have in society in the future, how do we want to include AI? Do we want it to augment our ideas? Do we want it to evaluate our ideas, or do we want it to replace jobs?
I would say that now we have an opportunity to actually talk to each other and use the collective intelligence of cities so we can finally combine expert knowledge with the knowledge of everyday life. We are able to bring together the technical with the experiential dimension.
What we're looking at with AI is not about changing your job, it is about starting to communicate with a bigger number of people, it's about expanding your platform.