What AI Costs the Environment

Generative AI is affecting the environment in more ways every day

As someone who cares about the environment and also makes a living from data science, the introduction of generative AI into everyday life in the last few years has made my life, well, complicated.

On the one hand, here is a new tool that could help us all make more discoveries with data — whether it’s through churning out code at unprecedented speeds (and helping folks nervous about coding more easily get involved!), synthesizing vast bodies of research, helping to iterate and refine hypotheses, collecting new data, or any of its many other capabilities. We’re already seeing AI help save lives through medication recommendations, allow researchers to better understand devastating diseases, and even predict and track the impacts of climate change, to name just a few exciting developments.

On the other hand, this tool comes with very real problems, including copyright infringement, built-in and reinforced biases, hallucinations, deepfakes, job loss in many sectors (with some evidence that humans are losing jobs to AI’s potential alone), harms to education and critical thinking, and never mind the looming specter of catastrophic, sci-fi horror-style outcomes from this technology (Chemical weapons! Disinformation campaigns! Rogue robots! Are we having fun yet?).

Of all of the concerns with AI, one of the most commonly cited is its impact on the environment. In fact, Americans are more likely to name AI as an environmental concern than other prominent environmentally harmful activities, including meat production, air travel, and cryptocurrency mining. One aspect that’s gotten a lot of attention in particular is the rapid growth of data centers, which are warehouses of computers used to train AI models. These centers are cropping up all over the country and the world, consuming massive amounts of energy, increasing local electricity rates, and even creeping into land set aside for our national parks.

So what can we do about all this? To answer that, it helps to first understand what we’re actually talking about.