Vault of Contemporary Art V&A Exhibition
At Make, we push the boundaries of creativity while ensuring our designs make positive contributions to the communities they serve. Our primary goal involves going beyond physical designs to craft spaces that transform people’s lives. For example, our virtual Vault of Contemporary Art gallery – a virtual project offering a parallel experience to a gallery space, where individuals from around the globe can freely enjoy exhibitions simultaneously. This initiative expands imaginative design horizons and is testament to Make’s creativity, providing avenues to enhance our projects by delving into new technology and possibilities.
Initial approach to AI
Make has been very proactive in the integration of AI in our design processes, with our founding partner, Ken Shuttleworth, recognising its potential early on. Alongside the existing sector groups we have at Make, we established a dedicated AI-focused specialism group, which serves as a hub for interdisciplinary collaboration and innovation. It’s important for us to bring together individuals from various departments to use the full potential of AI across the office. Establishing an AI Policy was an important step to indicate to all Makers the importance of exploring this field, and to provide guidelines, ensuring AI is used responsibly and ethically across all our projects.
Investment into research and internal training
Make has always made significant investments in internal training. Now more than ever, we understand the importance of continuous learning to equip our teams with the necessary skills and knowledge to leverage AI effectively. We have provided prompt-writing training sessions and, as expected, they have been popular across the studio and led to the development of an in-house prompting guide.
AI as a communication tool
To improve our communication, we have been using LLM models as a starting point for text and are working to develop a GPT that is consistent with our house style and voice. It’s used by our communications team as a tool to help us be more efficient in our communications. We’re employing the same tools in projects, creating project specific LLMs that can help us understand our briefs and project requirements by looking at the wider context of project information. Visual communication is aided by different diffusion models, helping to instantly generate ‘precedent’ images that convey the intended look and feel of a space.
Advancements and limitations
The potential of AI in architecture is significant, yet challenges persist in its full integration. In ongoing discussion with our visualisation team, a recurring challenge is the precise detail required and the difficulty to replicate an architect’s vision. AI aids in design, but it’s not a tool to produce project-ready outputs. However, it’s important to highlight the valuable role of AI in our workflow. Our visualisation team utilises AI to provide architects with visualisations of potential render set-ups, offering options for lighting, materials and more.
Our interiors team also uses AI to explore various design options based on their sketches, which they then present to the rest of the team before modelling them in Revit. Others have used AI to present specific space ideas to clients, such as staircase designs or material mood boards. I think it’s crucial to understand the main limitations with AI. Despite the valuable insights and inspiration that it can provide, the human touch remains indispensable in realising the full potential of architectural design.
AI tools
Our tech-focused team utilises and experiments with several AI tools to boost our design process. Tools like Midjourney, Stable Diffusion and LookX generate images from different inputs, including sketches, images and prompts, helping us to define and refine our design vision. Additionally, Forma by Autodesk is a versatile tool for generative design that allows us to explore myriad design options based on specific parameters. We use its near-real-time simulation tools for energy and daylight to assess the impact of potential design decisions.
In some cases, individuals within our teams are training models for Stable Diffusion based on our in-house project images or photographs. While we consider our freedom from a house style to be a strength of the practice, it has made the training process challenging. ChatGPT and its integration with DALL-E is widely used in our studio by all departments. In the BIM team, we particularly use AI as an assistant for developing custom scripts and code for tools tailored to the studio’s needs.
Despite the benefits that these tools offer, the significant challenge is the need for robust organised and systematised data inputs for training to ensure accurate outcomes. In addition, the ability to interpret and creatively assess the results requires the consideration and expertise of the design profession.
AI for sustainability
At Make, we’re prioritising the research of AI tools that specifically target the sustainability aspects for our projects. A tool we have started to use is Forma by Autodesk, which is useful in the initial stages of design, particularly for sustainable design considerations. While we may not have the resources to develop our own AI tools, we actively collaborate with leading companies that are experimenting with AI, especially those focused on sustainable solutions. While these tools are evolving, we can develop and future-proof our internal workflows to incorporate the tools when they become available.
The future
Looking forward, the integration of AI at Make will continue to evolve, presenting both new opportunities and challenges. A key opportunity for our BIM team is the reliable integration and interoperability of AI systems across various stages of the construction lifecycle. The current plethora of tools in development shows the various strands of research being undertaken. One area of concern for me is the management of sensitive project data between different AI systems, such as privacy and security data. It will be essential to establish robust standards and frameworks for the management of data as well as the description of geometry.
