What would the smartest city in the world look like? And how would we even begin to plan, build, and maintain it? This is exactly what Alphabet-owned Sidewalk Labs is trying to figure out with their latest Toronto waterfront development called Quayside. With the help of machine learning, and other enabling technologies, Sidewalk Labs is “reimagining cities to improve quality of life.” (Sidewalk Labs, 2018)
Beginning in the mid 1900’s, city planners began leveraging systematic observational data of human behavior to make planning decisions and inform city policy. (Depts.washington.edu, 2018) The insights garnered through this observation have been incredibly valuable in creating meaningful public spaces, managing traffic flows, and developing overall city planning solutions to support more effective and efficient interactions between people. Over the last 70+ years, however, there’s been little innovation in this space. While, today, we have pervasive behavioral monitoring technology throughout our cities such as infrared cameras, motion sensors, and geo-tracking, until recently it has been incredibly complicated to distill significant insights from the data collected by these technologies.
Sidewalk Labs is attempting to take a more calculated approach to solving city planning problems as they look to predict the societal impacts of design choices, and, ultimately, build cities that help residents live more connected and fulfilling lives. Sidewalk Labs is using the same theoretical components developed in the mid 1900’s around using observable human behavior to influence urban planning choices, but is supplanting the role of human observer with machine learning. The company’s charter for Quayside to “create a place that deployed emerging technology and people-first design innovations to address the challenges that face growing cities” (Sidewalktoronto.ca, 2018) is no small task, but one Sidewalk Labs has developed a robust framework to try and address.
The company has what they call a “digital layer” made up of four components; a city-wide distribution network of sensors, a map collecting location-based information about the infrastructure and resources, a personalized account portal for residents to access services, and a machine-learning powered model simulating various “what if” scenarios for city operations and long-term planning. Arguably the most complicated element of this digital layer lies in the machine learning component, because its success relies upon accurately predicting typically unpredictable customer behavior.
Sidewalks Labs is attempting to solve this issue beginning, in part, in the more straightforward and behaviorally consistent space of human travel modeling. The Model Lab team develops high-fidelity predictive models by creating a “synthetic population” statistically representative of a true population, training machine learning models to identify and eventually understand behavioral patterns, and then generate predictions for each person in the model as well as predictions around aggregate behavior for the synthetic population. Theories around human travel preferences, such as one which posits “travelers consider every minute waiting for a bus about twice as annoying as every minute riding on a bus,” (Sidewalk Labs, 2018) are tested and validated in these simulations, and the results of which grow the model’s so-called “intelligence” and accuracy. (Stories.dask.org, 2018) The company plans to use similar models to inform ideas related to land use, transportation, and government processes. Running these simulations is incredibly complex, but, as machine learning capabilities advance, they can be used to tackle more challenging urban planning questions around sustainability, mobility, safety and wellbeing. (Sidewalk Labs, 2018)
There are innumerable inputs that could result in vastly different results, which is where the true power of machine learning comes in. The technology has the capability to rapidly model out these scenarios with a reasonable degree of accuracy at a frequency humans alone never could. The more data the machine receives the more accurately complex situations can be modeled, considered, and ultimately predicted. But, while the outputs of this modeling are important, the greatest impact will come from a human analysis of the results to help account for a level of human-centered understanding a machine may never be able to reach. This balance between machine thinking and human analysis is a tricky one to strike, but in order to build the cities of the future they envision, Sidewalk Labs will need to master it.
While the opportunities Sidewalk Labs are exploring in this space are immensely exciting, there are two major questions I want to pose to the audience as we think about the future of machine-enabled city planning. First, as the physical world becomes more digital, how should we protect customer identity and data in smart cities? Second, how does technology change the culture of a city, and what are the negative consequences to consider in thinking about the impact on behavioral norms and interactions?
(Word count: 765)
Sidewalktoronto.ca. (2018). [online] Available at: https://sidewalktoronto.ca/wp-content/uploads/2017/10/Sidewalk-Labs-Vision-Sections-of-RFP-Submission.pdf [Accessed 10 Nov. 2018].
Depts.washington.edu. (2018). [online] Available at: https://depts.washington.edu/dmgftp/publications/pdfs/mouse_class/chapter2.pdf [Accessed 10 Nov. 2018].
Bond, M. (2018). The hidden ways that architecture affects how you feel. [online] Bbc.com. Available at: http://www.bbc.com/future/story/20170605-the-psychology-behind-your-citys-design [Accessed 8 Nov. 2018].
Garfield, L. (2018). Google’s parent company just reached an agreement with Toronto to plan a $50 million high-tech neighborhood. [online] Business Insider. Available at: https://www.businessinsider.com/google-sidewalk-labs-toronto-neighborhood-2018-7 [Accessed 3 Nov. 2018].
Sidewalk Labs. [online] Available at: https://www.sidewalklabs.com/ [Accessed 1 Nov. 2018].
Sidewalk Labs. (2018). Sidewalk Labs | A key to democratizing urban solutions is building better models. [online] Available at: https://www.sidewalklabs.com/blog/a-key-to-democratizing-urban-solutions-is-building-better-models/ [Accessed 1 Nov. 2018].
Stories.dask.org. (2018). Sidewalk Labs: Civic Modeling — Dask Stories documentation. [online] Available at: http://stories.dask.org/en/latest/sidewalk-labs.html [Accessed 1 Nov. 2018].