Using AI in EM

As emergency managers, should we be using generative AI?

AI is a dangerous tool. There are significant ethical and environmental considerations tied to its use. Data centres are often built in underserved, rural areas where residents have little control. There are concerns and complaints about noise and light pollution, unjust labour, habitat loss, and environmental degradation. The algorithm and how AI is trained is often biased, and results are not always accurate and are difficult to verify. AI is not neutral; returns are susceptible to discrimination and learned or inserted bias. Invention of AI-based products that quickly scan a person’s face and link to their identity, employment, home address, and other personal information, is alarming. There are significant public safety and privacy concerns with AI tools.

These exploitive concerns aside, AI is simply terrible for the environment. A few ChatGPT queries use approximately 2 cups of water. A single large data centre consumes about 1 million gallons per day. Our rivers and lakes are being bled dry. Power plants used to run data centres also use significant amounts of water. A single ChatGPT query needs approximately 420-milliwatt-hours of electricity, and about 10x more energy than a single google search. It is estimated that AI data centres are currently consuming about 1% of global electricity, and that number is rapidly rising.

As emergency managers, we work in spaces that are impacted by climate change. We see the devastation from longer and hotter fire seasons, increased flooding, droughts, more powerful storms and weird weather. By using generative AI, are we contributing to this cycle? Do we have a responsibility to avoid further damaging what’s already a damaged system? AI does have its uses – in non-generative, highly regulated situations. But generally, by using AI to create cute images and answer simple questions, I believe we are perpetuators of a harmful system.