AI’s growth in the field of architecture and design is inevitable but there are limitations to what it can do — such as replace human creativity.
Kostika Lala, principal and a founder of Toronto-based Flashcube Labs, a collective of designers and architects that embraces AI, said while it might represent a significant change to the industry, AI is in its infancy and design results “are still very basic.”
It won’t replace creativity, but it can be used to expand an architect’s creativity, he told a seminar audience recently on the impacts of the technology on architecture.
Flashcube employs AI for, among other things, the generation of renderings that can be quickly modified without having to model the project.
Lala was one of three panellists at the seminar held recently at The Buildings Show 2024 in Toronto.
While some detractors see AI taking away jobs, panellist Victoria Ikede told the audience it will also create new jobs. Employed by Architecture49 as the BIM co-ordinator, she said her job didn’t exist a decade ago.
She told the seminar at the Metro Toronto Convention Centre that because AI “is so good at text recognition, it really is a faster way of using and fine tuning the data from our BIM models.”
Training AI on a comprehensive data base provides the best results, she added, noting she is exploring AI using 3D models to get more out of coding.
But while it is best suited as a collaborator, not a search engine, it requires a lot of coaching and correction, Ikede said.
“It is like an intern who knows what it should do but has no idea when it is correct or knows the background behind what it does.”
Ikede said the technology has “highly randomized” output.
“You rarely, if ever, get the same output twice.”
Lala told the seminar audience what is frustrating for architects is that AI can hallucinate (generate information not based on input data) but there are methods of controlling it.
He said the biggest step forward for AI will be transitioning from images to 3D models.
He told the seminar the technology is suitable in academic environments as a research-based assistant for both students and teachers.
But panellist Indrit Alushani, a research associate at the University of Miami’s School of Architecture, said some universities don’t welcome the technology in their curriculum.
There will always be resistance to new technology and AI is no different, but Alushani believes architectural students should be given opportunities to understand it and leverage it in appropriate ways.
An exercise that Alushani assigns students is the creation of a multitude of AI iterations on a specific system, followed up with an examination of the various results.
To bolster security, Ikede said ideally design firms will train people on their own large language models (LLMs), rather than use open source AI such as ChatGPT or Gemini. LLMs also can be trained to the specific needs of the company.
Moderator Monifa Charles-Dedier, of ZAS Architects + Interiors, teaches architectural students at the University of Toronto’s Daniels School of Architecture Landscape and Design and at OCAD University.
She voiced concerns introducing AI too early in schools, before students learn to draw and make models can impede their learning process.
The panel also voiced concerns about AI’s impact on design copyrights. In 2022 Canada introduced the Artificial Intelligence and Data Act to ensure the design and deployment of AI is safe, transparent and non-discriminatory but Ikede said the act doesn’t get into specifics around ownership.
Copyright law is designed to protect work made by people, not machines, she said, adding it is a grey area because designers can be co-creating with AI.
Recent Comments