From homelessness to infrastructure deficits, to run-of-the-mill issues like snow removal, city councillors across Canada have full plates these days, and I’m willing to bet not many have artificial intelligence at the top of their agendas. But the AI Action Summit in Paris earlier this month highlighted the need for cities to consider AI when they are thinking about climate plans.
I attended the summit as a near-absolute beginner in the world of artificial intelligence. It was a rich learning environment with a varied slate of topics, from safety to governance.
I’m just starting to get my head around what the proliferation of AI means for local governments and their climate priorities. As a former city councillor in Saskatoon from 2011 to 2024, I will admit that AI’s impacts on the environment—and its implications for local governments more generally—were not something we discussed much around the council table.
What I learned at the summit, however, is that local governments should be thinking about AI in their considerations of climate, water, and energy use in a few different ways.
First, local governments should consider how AI can optimize operations and improve efficiency for municipalities. As we heard throughout the AI summit, although there are environmental costs to using AI, there are also costs to be borne by neglecting to do so. If they choose not to take advantage of tools available to analyze data and improve operations, local governments and organizations will leave money on the table—which explains why many if not most Canadian cities are already engaging in this work to some degree.
At the same time, a second take-home message from the summit is that when it comes to accurately measuring AI’s environmental implications, there is very little consensus to be found. There are varying perspectives on the severity of this issue, with some voices signaling the risk of AI’s negative environmental impacts being distributed inequitably and others urging caution in assessing the energy and water demands of AI.
There are important efforts under way to standardize this kind of analysis, including the introduction of AI Energy Scores, which is meant to be something like an EnergyStar equivalent for AI models. Many experts also spoke of the need to prioritize “frugal AI”, models that match the tool to the task rather than risking computational “overkill”’ as a means of limiting environmental consequences.
This is an optional title









