In this interview, Pál Mészáros delves into the transformative role of artificial intelligence in architecture, exploring its impact on design methodologies, the evolving relationship between human creativity and computation, and the enduring role of architects in an AI-driven future. As a developer of cutting-edge software for area optimization, floor plan generation, and procedural design, Mészáros offers a forward-thinking perspective on how these advancements are reshaping the profession.
This discussion sets the stage for the BIG Architecture festival in Portorož (May 22–23), where Mészáros will take part in a debate on AI’s role in shaping the future of architectural practice.
Zala Zia Lenárdič: In recent years, artificial intelligence has moved from theoretical frameworks to practical applications across nearly every field—including architecture. AI-driven design agents, once limited to generative models based on finite constraint principles, have now evolved into complex, interconnected neural systems capable of producing remarkable architectural solutions. These advancements not only challenge traditional workflows but also redefine the architect’s role in an era where machines contribute to the creative process. With this in mind, could you share how your practice integrates AI into architectural design, and what excites you most about its potential to shape the future of the profession?
Pál Mészáros: I think we can showcase what we’ve been working on over the past six months or so.
Yes, I sent you something. Maybe I should start with the first project—the Plan File plugin. Or would you prefer I begin with something else?
Recently, we have been developing a web application. You mentioned it as an agent that can assist in our work, though that might be a bit of an overstatement at this stage. Right now, it functions more as a manual application, allowing users to create a site anywhere—for example, on an empty plot—and quickly draft a warehouse layout.
This is our site, and as you can see, it includes a grid since these buildings are typically arranged in a structured format. It’s very easy to start drawing a building, and the system automatically generates a surrounding road. The building can be rotated and moved within the site as needed.
After that, docking areas can be added where trucks can attach to the building, and the system adjusts the road accordingly. We are planning to introduce many more features. For example, we can currently add multiple buildings to a single site, but the tool is still in its demo phase. The core concept is quite similar to our previous project, Plan File. However, we faced challenges with Grasshopper—it is a highly complex tool within Rhino, functioning as a plugin inside another plugin, making it difficult to explain and use effectively.
The main idea remains the same: to automate parts of the design process. While automation may reduce certain aspects of creativity, it significantly enhances control over beneficial design outcomes. It allows for rapid iteration through multiple options and enables analysis using tools such as daylight autonomy, energy simulations, and spatial calculations, including room counts and total square footage.
Regarding the AI agent concept, I view it more as a sophisticated algorithm that assists in design rather than true artificial intelligence. However, if machine learning were integrated into this process, we could automate more aspects of design by leveraging user-generated data. This could enable the system to generate new designs automatically. That is the vision. I am not sure it will be that straightforward, but at this stage, it functions more like an advanced algorithm than a fully autonomous AI system. Our approach is to develop algorithms that generate buildings, with the eventual goal of having this process guided by an AI agent. This is the direction we are moving toward.
Additionally, we incorporate AI-driven image creation into our workflow. This is just one of the ways we use AI. Beyond design automation, we utilize AI for generating concepts, producing text, creating sketches, and assisting at various stages of the architectural process. For systematic buildings that follow a grid pattern, simple algorithms can efficiently generate repetitive structures. However, when it comes to public buildings, the approach is entirely different. AI is not yet advanced enough to effectively generate floor plans for such structures.
“If AI alone were designing buildings, future architecture might be optimized for other AI agents rather than humans!”
That said, we have experimented with AI-generated facades and other exploratory methods. AI provides a rapid way to test ideas and visualize concepts before committing to a full design. We also developed a tool called Visual AI, which follows a similar approach. It is a mobile and web-based application that generates renders from simple sketches. For example, if you provide a rough sketch of an apartment in an eye-level view, the AI can create a highly realistic render based on the input prompt. The control strength feature allows users to determine how closely the generated image should resemble the sketch or whether it should take a more creative direction.
This tool significantly speeds up the traditionally time-consuming rendering process. Traditional rendering requires substantial effort, and if not executed well, it can result in poor representations of design concepts. By leveraging AI, architects can quickly generate multiple variations of an image, refine their vision, and iterate more efficiently before finalizing their designs.
Tomaž Šimnovec: This brings us to an interesting question—what happens if we replace every part of the architectural process with AI? What would that mean for small studios like yours? How would it impact the field as a whole?
Pál Mészáros: Well, if AI alone were designing buildings, future architecture might be optimized for other AI agents rather than humans! It’s a complex issue—if too much of the process is automated, does architecture retain its human-driven essence, or do we gain an entirely new way of thinking? The question of whether AI should fully take over design is a difficult one.
Humans learn through experience, starting from childhood. The way we perceive and interact with spaces is shaped by our personal environments. AI, on the other hand, processes information differently—it weighs data equally, without the emotional and contextual nuance that human designers bring.
However, AI can be a powerful tool when used correctly. Rather than replacing architects, it could serve as an advanced assistant, allowing designers to focus on higher-level creative and conceptual work. This could foster a new form of collaboration between human intuition and machine efficiency.
Another key consideration is that architecture is a multi-sensory experience. Spatial perception extends beyond visuals to include textures, smells, temperature, and atmosphere—elements AI currently cannot fully comprehend or manipulate. For now, AI is best suited as a visualization and communication tool rather than a complete replacement for human designers.
“Rather than replacing architects, AI could serve as an advanced assistant, allowing designers to focus on higher-level creative and conceptual work. This could foster a new form of collaboration between human intuition and machine efficiency.”
Tomaž Šimnovec: Just a follow-up question—have you ever been truly surprised by an AI-generated solution? Has AI ever produced something unexpected or groundbreaking in your work?
Pál Mészáros: Yes! We once participated in an architectural exhibition in Hungary where we used Visual AI to generate images. Some of the results were truly unexpected. For example, some designers have experimented with using images of fruit to inspire skyscraper designs, which was a fascinating approach we had not considered before.
For me, the most surprising AI experience was more technical. I needed to calculate a rotation using trigonometric functions. Instead of figuring it out manually, I asked ChatGPT, and it gave me a correct script instantly. While I did not fully understand the code at first, it worked perfectly. However, I later discovered an issue in the code, highlighting that while AI can be useful, human oversight is still necessary.
Curators: As a final thought, what are you looking forward to in the AI debate?
Pál Mészáros: I am interested in how AI will shape careers in architecture, especially in emerging fields like prompt engineering. It will also be fascinating to see how different professionals adapt AI to their workflows and how this influences the future of design.
Yes, and I hope AI remains a tool that enhances our work rather than a force that replaces creativity altogether.
In this interview, Pál Mészáros delves into the transformative role of artificial intelligence in architecture, exploring its impact on design methodologies, the evolving relationship between human creativity and computation, and the enduring role of architects in an AI-driven future. As a developer of cutting-edge software for area optimization, floor plan generation, and procedural design, Mészáros offers a forward-thinking perspective on how these advancements are reshaping the profession.
This discussion sets the stage for the BIG Architecture festival in Portorož (May 22–23), where Mészáros will take part in a debate on AI’s role in shaping the future of architectural practice.
Zala Zia Lenárdič: In recent years, artificial intelligence has moved from theoretical frameworks to practical applications across nearly every field—including architecture. AI-driven design agents, once limited to generative models based on finite constraint principles, have now evolved into complex, interconnected neural systems capable of producing remarkable architectural solutions. These advancements not only challenge traditional workflows but also redefine the architect’s role in an era where machines contribute to the creative process. With this in mind, could you share how your practice integrates AI into architectural design, and what excites you most about its potential to shape the future of the profession?
Pál Mészáros: I think we can showcase what we’ve been working on over the past six months or so.
Yes, I sent you something. Maybe I should start with the first project—the Plan File plugin. Or would you prefer I begin with something else?
Recently, we have been developing a web application. You mentioned it as an agent that can assist in our work, though that might be a bit of an overstatement at this stage. Right now, it functions more as a manual application, allowing users to create a site anywhere—for example, on an empty plot—and quickly draft a warehouse layout.
This is our site, and as you can see, it includes a grid since these buildings are typically arranged in a structured format. It’s very easy to start drawing a building, and the system automatically generates a surrounding road. The building can be rotated and moved within the site as needed.
After that, docking areas can be added where trucks can attach to the building, and the system adjusts the road accordingly. We are planning to introduce many more features. For example, we can currently add multiple buildings to a single site, but the tool is still in its demo phase. The core concept is quite similar to our previous project, Plan File. However, we faced challenges with Grasshopper—it is a highly complex tool within Rhino, functioning as a plugin inside another plugin, making it difficult to explain and use effectively.
The main idea remains the same: to automate parts of the design process. While automation may reduce certain aspects of creativity, it significantly enhances control over beneficial design outcomes. It allows for rapid iteration through multiple options and enables analysis using tools such as daylight autonomy, energy simulations, and spatial calculations, including room counts and total square footage.
Regarding the AI agent concept, I view it more as a sophisticated algorithm that assists in design rather than true artificial intelligence. However, if machine learning were integrated into this process, we could automate more aspects of design by leveraging user-generated data. This could enable the system to generate new designs automatically. That is the vision. I am not sure it will be that straightforward, but at this stage, it functions more like an advanced algorithm than a fully autonomous AI system. Our approach is to develop algorithms that generate buildings, with the eventual goal of having this process guided by an AI agent. This is the direction we are moving toward.
Additionally, we incorporate AI-driven image creation into our workflow. This is just one of the ways we use AI. Beyond design automation, we utilize AI for generating concepts, producing text, creating sketches, and assisting at various stages of the architectural process. For systematic buildings that follow a grid pattern, simple algorithms can efficiently generate repetitive structures. However, when it comes to public buildings, the approach is entirely different. AI is not yet advanced enough to effectively generate floor plans for such structures.
“If AI alone were designing buildings, future architecture might be optimized for other AI agents rather than humans!”
That said, we have experimented with AI-generated facades and other exploratory methods. AI provides a rapid way to test ideas and visualize concepts before committing to a full design. We also developed a tool called Visual AI, which follows a similar approach. It is a mobile and web-based application that generates renders from simple sketches. For example, if you provide a rough sketch of an apartment in an eye-level view, the AI can create a highly realistic render based on the input prompt. The control strength feature allows users to determine how closely the generated image should resemble the sketch or whether it should take a more creative direction.
This tool significantly speeds up the traditionally time-consuming rendering process. Traditional rendering requires substantial effort, and if not executed well, it can result in poor representations of design concepts. By leveraging AI, architects can quickly generate multiple variations of an image, refine their vision, and iterate more efficiently before finalizing their designs.
Tomaž Šimnovec: This brings us to an interesting question—what happens if we replace every part of the architectural process with AI? What would that mean for small studios like yours? How would it impact the field as a whole?
Pál Mészáros: Well, if AI alone were designing buildings, future architecture might be optimized for other AI agents rather than humans! It’s a complex issue—if too much of the process is automated, does architecture retain its human-driven essence, or do we gain an entirely new way of thinking? The question of whether AI should fully take over design is a difficult one.
Humans learn through experience, starting from childhood. The way we perceive and interact with spaces is shaped by our personal environments. AI, on the other hand, processes information differently—it weighs data equally, without the emotional and contextual nuance that human designers bring.
However, AI can be a powerful tool when used correctly. Rather than replacing architects, it could serve as an advanced assistant, allowing designers to focus on higher-level creative and conceptual work. This could foster a new form of collaboration between human intuition and machine efficiency.
Another key consideration is that architecture is a multi-sensory experience. Spatial perception extends beyond visuals to include textures, smells, temperature, and atmosphere—elements AI currently cannot fully comprehend or manipulate. For now, AI is best suited as a visualization and communication tool rather than a complete replacement for human designers.
“Rather than replacing architects, AI could serve as an advanced assistant, allowing designers to focus on higher-level creative and conceptual work. This could foster a new form of collaboration between human intuition and machine efficiency.”
Tomaž Šimnovec: Just a follow-up question—have you ever been truly surprised by an AI-generated solution? Has AI ever produced something unexpected or groundbreaking in your work?
Pál Mészáros: Yes! We once participated in an architectural exhibition in Hungary where we used Visual AI to generate images. Some of the results were truly unexpected. For example, some designers have experimented with using images of fruit to inspire skyscraper designs, which was a fascinating approach we had not considered before.
For me, the most surprising AI experience was more technical. I needed to calculate a rotation using trigonometric functions. Instead of figuring it out manually, I asked ChatGPT, and it gave me a correct script instantly. While I did not fully understand the code at first, it worked perfectly. However, I later discovered an issue in the code, highlighting that while AI can be useful, human oversight is still necessary.
Curators: As a final thought, what are you looking forward to in the AI debate?
Pál Mészáros: I am interested in how AI will shape careers in architecture, especially in emerging fields like prompt engineering. It will also be fascinating to see how different professionals adapt AI to their workflows and how this influences the future of design.
Yes, and I hope AI remains a tool that enhances our work rather than a force that replaces creativity altogether.
Curators:
Tomaž Šimnovec & Zala Zia Lenárdič
Powered by
