When Zach Katz looks at the world around him, his artist's eye enables him to imagine a "more beautiful" version of his surroundings.
But rather than paint, pencil or clay, he is using emerging artificial intelligence tools to help bring the enhanced versions to life.
The founder of BetterStreetsAI started off using image-based artificial intelligence programs to transform his own New York neighbourhood.
"I started with photos of my street here in Brooklyn and made it into a pedestrian promenade, added a water fountain and some kids playing and it just looked really nice," he said.
"When I look at a street I can always see how much more beautiful it could look, and how much more people-centric and less car-centric it could be.
"It was cool to easily be able to visualise that with technology."
Katz does not have an urban design or urban planning background, but said he had picked up "good 21st-century urban planning principles", in part from a stint living in Amsterdam.
"It's not rocket science. Just take away space from cars and all of a sudden you get a good street".
Once Katz starting posting the street transformations online, he began receiving requests from people all over the world to makeover their neighbourhoods.
"I got a couple of thousand requests, but ultimately I just do whatever I feel like doing," he said.
"So I'll just go to Google Maps, pick a city and then click around until I see a street that looks like it would be fun to transform."
Using AI tools 'an art form in itself'
Among Katz's transformations have been streets in Sydney's Marrickville, Brunswick in Melbourne and Brisbane's West End.
He uses a version of DALL-E, an AI system described as a "sibling" to ChatGPT that can generate or alter imagery based on language inputs.
Katz said transforming streets required a fair bit of work on his part.
"It's a bit of a process, just like making any art, and also about having a vision for how you think it should look to sort of guide the AI in the direction you want," he said.
He said the current process involved uploading an existing picture of a street, erasing certain parts of the scene and then using prompts to guide the program to fill in the gaps.
Katz said learning how the AI program would respond to particular prompts was "a bit of an art form in itself" to enable him to "additively paint a picture".
"The phrase 'car-free street' and the phrase 'pedestrian promenade' can yield very different results," he said.
"Most of my learning curve was accustoming myself to the nuances of the results these different phrases yield."
"Sometimes you can trust the AI to just come up with something good but usually it yields better results if you're more intentional about it."
How AI tools can help adapt urban environments for climate change
Hank Haeusler, an associate professor at the University of New South Wales' arts, design and architecture faculty, said AI did have an important role to play in urban design in Australia, but in a different way to that suggested by Katz's project.
He described the project as "only a visual art practice" with limited usefulness in real-world planning.
"What we see here is just a collection of popular items that you could put onto the street, such as putting a garden there, or more greenery around, or a cycle path, or you change the pavement to make it pink, and more appealing," he said.
"These are all good ideas but there are some fundamental problems it would face in Australia in terms of what our streets need."
Dr Haeusler, who also heads up the university's AI institute, said the biggest challenge facing Australian cities was climate change.
He said nuanced AI technology could be used to help adapt streets to better cope with the results of extreme weather events such as heatwaves and flooding.
"To actually transform the streets is a very, very long and difficult undertaking and we're running out of time," he said.
"We had floods over the past few years, we had bushfires, we had extreme heat."
He said addressing those issues in major cities could mitigate the effects of climate change.
"Quite often the cities are responsible for urban heat, [with] too much concrete creating urban heat islands," he said.
"For floods, if every surface is concrete then water can't penetrate into the ground and therefore gets flushed down into sewage, into the river, and the river swells and floods the area."
Increasing greenery in urban environments wherever possible was often key to addressing these issues, Dr Haeusler said.
He said the process of designing streets to help mitigate heat and flooding issues could be sped up using increasingly advanced computational tools.
"An urban designer, a council could look at the existing street and say ... we've got a heat profile of X, the penetration level is Y and the green rating is seven," he said.
"We can create a baseline condition, then out of that baseline condition you can put a tree in there and see what kind of impact it has on heat, and greenery.
"You could say let's change the material from concrete to another material, and see what kind of impact that has on cooling."
He said such AI-enabled tools could enable those designing the streets to get "very, very rapid" information to inform their decisions.
"You could design the street and redesign the street ... assessing [whether] your design has a certain impact towards those sustainable measurements," he said.
He said without such technology, making decisions about how and where to swap out concrete or plant trees in specific locations could be a very lengthy and time-consuming process for planners.
"You need something that can help you shorten that length so you can make informed decisions very, very rapidly, by having all the information available. That's where AI can help you," Dr Haeusler said.
For Katz, his street transformations use technology as a "means to an end" to inspire people to imagine the best versions of public spaces.
"Whether it's government officials or urban planners using it, or activists or artists — my goal is just to show people how much better the streets could be," he said.
"It's very obvious looking at the before and afters, which one looks better.
"Which one looks like a place you'd rather live?"
