The challenge in using artificial intelligence to build the city of the future is that, well, the city is already there. Unlike in the video game SimCity, where players build from scratch, there’s existing infrastructure that needs to be taken into consideration.

New York City, Mexico City, Paris, even Pittsburg – these cities are literally set in stone. Neither the finances nor the willpower exist to tear down buildings, rip up existing roads, and start fresh. So, as happened in the past in the evolution of laying cobblestones over dirt and then paving the cobblestones, AI developments will need to be layered neatly on top of the existing structures to work.

And, in fact, that’s what several companies are already working toward. From using sensor data that automatically generates adjustments in traffic and transmission grids, to automated modeling that helps cities smartly deploy resources, companies are lining up to find ways to use AI to enhance and improve cities around the globe.

Academics are also tracking these technology developments, which give them new ways to take their findings about cities and quantify them for use by planners and architects. For example, the Collective Learning Group at the MIT Media Lab is now using crowdsourced data combined with machine vision to analyze changes that cities have already made.

By closely examining neighborhood changes or improvements, Cesar Hidalgo, the Collective Learning Group’s director, and his team are able to examine the theory behind a planned change, the change itself, and the outcome. This will allow cities to discard disproven theories and intelligently deploy increasingly limited resources.