A collage of the student's faces and their work

A piece of AI called DALL-E can make pictures based on what you write

Khalil Heron says that when he was asked to make a plan for a more environmentally friendly university campus, he wanted to make “a perfect utopia.”

Katherine Perrott, the professor of the second-year design studio foundations class at the University of Waterloo, told the student to use an artificial intelligence (AI) system called DALL-E to start making plans.

DALL-E can make pictures based on written instructions. The system was made by the same person who made ChatGPT, a chatbot that has been getting attention lately for being able to make conversations sound like they came from a real person.

Heron wanted to see what would happen if he made the final picture look more like a painting or drawing.

“I wanted it to have a solarpunk style, which is almost like cyberpunk but puts a lot of emphasis on being green.” “I wanted to build a perfect utopia where everything could last forever,” he said.

“I think it’s amazing that the program can make not only real images but also works of art.”

The focus of his final design was green transportation and urban spaces.

‘A visual brainstorming tool

Perrott said that she told the students to use ideas from the university’s own sustainability reports and course readings to figure out what keywords to put into the system.

She said, “AI tools are so new, and they are interactive and quick, which makes them a lot of fun for students to use.” “I really think of it as a visual way to come up with ideas, like sketching.”

Perrott says that the purpose of the exercise was to show students how AI can be used early on in the design process to come up with ideas and allow “creative exploration” to quickly turn ideas, concepts, and key words into visuals.

woman with curly hair and blue striped shirt smiling

She said, “The next step will be to do more traditional exercises with graphic design software, rendering, and modeling to come up with ideas that are more relevant to the University of Waterloo.”

Making changes to get the right pictur

A second-year student named Myah Sachedina said that trying out different keywords and prompts helped her figure out which ones worked best for her vision.

“I started with a tall, glass, futuristic building with a greenscape and a sustainable bullet train on a bridge,” she said, explaining that her idea was based on sustainability. “But the image that came out didn’t really match what I wanted, so I had to add more commands and change some of the wording.”

blue and green city scape with people sitting by a green hill with a glass bridge above them

She said that she added “photorealistic” as a prompt to avoid getting results that looked too cartoony.

She got ideas for her final design from the green spaces and modern buildings she saw when she was growing up in Toronto.

Red, white and yellow image of an urban landscape, with a red and white bus at the centre of the picture and people walking and bicycling around it.

Students like Sachedina and Heron found that they had to be very clear about what kind of image they wanted the AI system to make.

Perrott said that there has been a lot of discussion about artificial intelligence and academic integrity because AI tools use ideas and images from already-published work on the web.

But for this project, students are only allowed to use AI as a tool, and they have to cite it in any work they turn in for a class project that uses AI.

Check out the other pictures that students have made here:

green space with a grey building full of plants and people walking on a sidewalk in the park1 of 6