The dark side of urban artificial intelligence: addressing the environmental and social impact of the algorithms

Artificial Intelligence has become one of the biggest hype topics of the year, also in the urban world. And not without good reason: AI is widely perceived as a valuable tool to address the most important challenges of our societies, including dealing with the climate emergency. The potential uses of AI applications are extensive and varied, ranging from enhancing resource usage efficiency, improving transportation and urban planning effectiveness, or providing more personalized urban services. AI is also the driving force behind the green and digital transition promoted by cities worldwide. Yet, while we tend to imagine the digital as something ethereal, limitless, dematerialised and neutral, there are significant social and environmental costs associated with the growing reliance of our societies on AI systems.  

To start with, the computing capacity needed to develop and deploy AI systems requires large data infrastructures, especially data centers that consume vast amounts of energy and water. To put it in numbers: data centers account for 1,5% of global electricity use, and training just one AI model can emit more carbon dioxide than the lifetime emissions of an average car. As these energy-hungry infrastructures quietly move into urban centers, it becomes imperative to make them more green and efficient if cities are to meet their climate commitments. In turn, AI hardware (chips and circuit boards) depends to a large extent on rare metals and rare earth, whose extraction entails important environmental, social and geopolitical costs. Finally, the need for regular hardware replacement can lead to increased material requirements and large amounts of electronic waste.  

On the social front, many of the algorithmic systems used by local governments to support decision-making are prone to incorporating and reinforcing our existing societal biases. Addressing them requires adopting a digital-rights perspective and ensuring that its deployment considers basic ethical principles such as non-discrimination, transparency, privacy protection, or accountability. A related fundamental question is how algorithmic curation may influence the character and quality of our democracy. Lastly, from a global justice perspective, much of the AI innovation used by cities worldwide is powered by underpaid workers in the global south. 

The seminar methodology 

CIDOB’s Global Cities Programme, in collaboration with Barcelona City Council, organises an international seminar to reflect on the dark side of the extensive use of artificial intelligence from an urban perspective. Specifically, it seeks to improve our understanding of the environmental and social negative externalities that the development of AI systems entails and to discuss how to mitigate or overcome them through specific local initiatives. 

The programme consists of two blocs. Each bloc opens with a 10-minute intervention by two keynote speakers, followed by a moderated session around a set of core questions that will be circulated in advance. To foster a dynamic debate and new insights, the moderator may call on individual participants to intervene at a relevant point of the discussion. The seminar's conclusions will contribute to the research of the Global Observatory on Urban Artificial Intelligence, and inform a policy paper to be published by CIDOB in the coming months.