Cities as Data Machines

Imagine waking up attached to a device that, powered by AI, continuously monitors everything you do: Where you go and how, when you walk, drive, cycle, or take public transport, how many steps you take, what you read, watch, and listen to, and even monitors your vital signs, your heart rate, your oxygen levels, and every one of your breaths. We all know these devices very well and willingly carry them around and wear them so that they can collect enormous amounts of data about us and offer all sorts of opportunities to “optimize” our lives.

Now imagine a city that monitors everything that composes its existence: The movement of people and goods, the energy exchanges of its infrastructure, every new or adapted construction, its natural systems, the vegetation, water, and other species that live in it, and the program and use of each of its public spaces. This is so far from the reality of even the most technologically advanced cities that it almost sounds like a dystopia.

The imminent and probably inevitable advent of the widespread disruption that artificial intelligence (AI) will bring to every aspect of human societies naturally brings along pressing questions to architects, planners, and all other decision-makers of cities. How will AI change the way we perceive, conceive, construct, and govern our urban environments? What role will data play in the evolution of cities, buildings, and the professions that govern them? At the end of the day, how is it possible that tech giants know more about how our cities operate than city planners and architects?

According to the World Economic Forum, at our current rate, we generate 2.5 trillion gigabytes of data every day. A smart city of about one million people generates two hundred million gigabytes per day. Fifty percent of this data comes from Internet of Things (IoT) devices, while more than 10% comes from social media. However, less than 1% of this data is utilized to foster innovation and drive decisions. At the same time, cities occupy less than 2% of the world’s land and consume 80% of its energy. The untapped potential for optimizing our cities is evident. And AI offers a groundbreaking opportunity to unlock these untapped resources.

This is why the growth of AI in the architecture and planning field has been explosive. In the past five years, our work at the Institute for Advanced Architecture of Catalonia (IAAC) has gone from a few seminars and studios to the birth of the Master in AI of Architecture and the Built Environment. The growth of the field is so rapid that we can’t reuse the curriculum we teach from year to year. From the first generative adversarial network (GAN) models that envisioned floor plans to the Stable Diffusion explosion that shattered architectural visualization to today’s large language mode (LLM) and retrieval augmented generation (RAG) models that blur the lines of artificial and human intelligence, it’s hard to keep up.

In more practical terms, it’s possible to teach architects and planners with zero AI experience to employ complex machine learning and generative models to revolutionize the analysis and design of space. To name a few examples, they range from simple clustering algorithms that balance the energy usage and potential production in a city to identifying the development potential of urban voids or image segmentation models that use simple street view images to generate accurate maps of street conditions for safe biking or map human activity in a city.

Today, AI software startups are booming in architecture and planning, expanding the creativity of designers to imagine dreamy landscapes and unprecedented forms. At the same time, our commitment as architects and planners to respond to the impact that our constructions have on the climate remains the same and AI provides a unique opportunity to achieve that.

More than ever, now it’s possible to integrate climate science, advanced technology, and design knowledge to make our buildings and cities more sustainable. At infrared.city a software company that was born from our research at the Austrian Institute of Technology, we are working to bring climate simulations available to every architect. Using cutting-edge deep learning technology and large simulation datasets, we train AI models to predict results and provide instant feedback on design decisions. A wind simulation that used to take a week is now instantly available at the click of a button. Combining these instant simulation results with extensive knowledge bases and language models allows us to bring the knowledge and simulation skills of a climate expert to every architect and planner.

Of course, this unique opportunity to exponentially grow our design capabilities through AI comes with significant responsibility. Just as computer-aided design (CAD) disrupted and fundamentally changed the architecture profession, AI has the potential to revolutionize what we understand, visualize, and build in the world. However, it cannot change our intentions and if AI technology is the answer, then we must certainly ask the right questions.

Main image: Climate simulations dataset for infrared.city’s AI models. Image © infrared.city