Document Type

Thesis, Senior




Spring 2019


smart-city, Quayside, Toronto, google A.I., machine learning, architecture, digital infrastructure






How intelligent is the typical smart-city design approach? In an era when artificial intelligence and big data promise to improve urban life in unprecedented ways, are smart cities being imagined and designed in ways that are actually inspiring and truly innovative? This project examines the proposals and approaches of Sidewalk Labs’ designs for Quayside, located in Toronto Canada, and asks how intelligent, really, is the city the propose. Sidewalk Labs (the city-building subsidiary of Alphabet, Google’s parent company), in partnership with Waterfront Toronto (a government-appointed nonprofit development corporation), claims to be reimagining cities from the internet up.” But is their project more than a corporate optimization of the usual “smart” themes of sustainability, data collection, efficiency, economic development, and technology?

A Smart Intelligent City poses the question, how could Google A.I. and Machine Learning technology, specifically Google Cloud Vision, be utilized as a design tool and source of design material to enhance or alter the conventional design and planning process used by Sidewalk Labs for Sidewalk Toronto at Quayside?

An Intelligent Smart City utilizes Quayside’s current technologies and digital infrastructures to dynamically generate new visual environments within the built environment with the use of Google Cloud Vision technology as the foundation to speculate on a similar but different algorithm, an architectural one.

The algorithm detects “architecture” from the ubiquitous influx of image uploads and stored into the Google Cloud by the individuals residing within the Quayside community, in addition, the algorithm also detects “architecture” within images that are searched through keywords and faceted navigation, a typical Google image search. Once detected, the algorithm then explores the architectural images, analyzing and examining the images’ contents, attributes, format, etc. to determine if the images qualify to move forward in the design process – this examination and approval of images is done in collaboration with the architect.

Finally, the algorithm operates on a collection of qualified/approved images to generate a “new architecture.” Some of the operations that function within the algorithm consist of blend, collage, merge, stylize, crop, filter, etc. The new architecture is then generated in the form of a watch, incorporating an assemblage of images that contain an array of architectural elements. The algorithm is programmed to resemble methods and techniques in which architectural images are produced by contemporary architects – through a series of operations incorporated within architectural, engineering, animation, etc. computing and design software.

As the algorithm continues to generate swatches over time, it learns which types of images generate the best swatches. The city visually responds to its complex environment, it’s multi-layered system, by processing an additional layer from its inhabitants – their image activity. Architecture in the city is no longer fixed; it is intelligent enough to constantly generate its own image.

Additional Information

Thesis Advisors:

Mark Linder


Local Input

Creative Commons License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial-Share Alike 4.0 International License.

Included in

Architecture Commons