The integration of Artificial Intelligence (AI) in architectural design is rapidly evolving, yet experts agree that AI is far from replacing the core human element of creativity. Despite the growing use of AI in the industry, professionals emphasize that the technology still has limitations that prevent it from fully replicating the nuanced and imaginative work done by human architects.

AI in Architectural Design: A Tool for Enhancement

Kostika Lala, founder of Flashcube Labs in Toronto, argues that while AI has already started transforming the architectural design landscape, its current capabilities are still rudimentary. According to Lala, AI-generated designs are 'quite basic' at the moment. However, he sees AI not as a replacement for creativity, but as a tool that can enhance the work of architects. His firm is using AI to generate renderings, which can be swiftly modified without the need for re-modeling, allowing architects to explore a wider range of possibilities in less time.

At the 2024 Architecture Expo in Toronto, experts discussed the implications of AI's role in the industry. Victoria Ikede, a BIM coordinator at Architecture49, noted that while some fear AI may take away jobs, it will actually create new opportunities. She pointed to the example of the BIM coordinator role, which did not exist ten years ago but is now a key position in architectural design.

mage Source Note: Image generated by AI, image licensed from Midjourney

AI as a Collaborative Tool in Design

During the seminar, Ikede highlighted AI's strength in text recognition, which can streamline the process of working with Building Information Modeling (BIM) systems. AI can help architects manage data more efficiently, enabling faster fine-tuning of models. However, she emphasized that AI is not a replacement for human expertise. Instead, it acts as a collaborative tool that requires substantial guidance. She likened AI to an intern who understands the task but lacks the depth of understanding to execute it perfectly. One of the key challenges with AI, according to Ikede, is that its output is often unpredictable: 'You can almost never get the same output twice.' This unpredictability, while useful in some contexts, can be frustrating for architects trying to finalize designs.

Lala also spoke about AI's tendency to generate 'hallucinations'—information that is not based on any input data—which can lead to inaccurate or irrelevant results. He sees this as one of the biggest challenges in the current phase of AI in architecture but believes that better control over these outputs is possible with further development.

AI's Potential in Academia and Industry

AI is already being used in educational settings, with Lala suggesting that it could serve as a valuable research assistant for students and teachers. However, not everyone in academia is ready to embrace the technology. Indrit Alushani, a researcher at the University of Miami's School of Architecture, pointed out that some universities are resistant to incorporating AI into their curricula. Alushani believes that students need the opportunity to understand AI's potential and limitations, rather than being kept away from it due to initial resistance. As part of his teaching, Alushani asked students to generate multiple AI iterations for a design system and analyze the outcomes to better understand the technology's capabilities.

Security and Ethical Considerations

As AI continues to gain traction, there are concerns about its impact on security and intellectual property. Ikede recommended that architectural firms train their own large language models (LLMs) rather than relying on open-source tools like ChatGPT or Gemini. Custom LLMs can be tailored to meet the specific needs of a firm, potentially offering a more secure and efficient solution.

Monifa Charles-Dedier from ZAS Architects + Interiors raised concerns about introducing AI into schools too early, suggesting that students should first master the basics of drawing and modeling before relying on AI. She expressed concern that prematurely incorporating AI into curricula might hinder students' learning processes.

Finally, the panel discussed the legal implications of AI in design, particularly around copyright. While Canada's Artificial Intelligence and Data Act, passed in 2022, aims to ensure AI deployment is safe and transparent, it does not explicitly address the ownership of AI-generated designs. Since AI often collaborates with human designers, the issue of copyright becomes murky. Ikede noted that current laws are designed to protect human creators, not machines, creating a gray area when it comes to co-creations between humans and AI.

Key Points

1. AI in architectural design is evolving but cannot replace human creativity.
2. AI enhances creativity by assisting with tasks like rendering and data management.
3. AI's unpredictable output and 'hallucinations' present challenges for architects.
4. AI's role in education is growing, but some institutions resist its integration into curricula.
5. Security and legal issues around AI-generated designs need further consideration.