AI is booming. The global artificial intelligence market is advancing day by day, with calculations expecting it to grow from the billions of dollars it is worth now to trillions in the next year alone.

Just about every major conglomerate has already invested in artificial intelligence: Microsoft, Meta, Google, Amazon, and more. We’re already seeing ways that the commercial and personal use of AI is shaping our culture — from how we think to what is considered art. But one of the major concerns people are raising about AI is its effects on the environment.

“It’s naturally concerning that something that’s already become as integrated into our daily lives as AI will have an even bigger environmental impact as it grows.”

Despite scientists ringing alarm bells for decades, the devastating effects of climate change are beginning to take hold —  so it’s naturally concerning that something that’s already become as integrated into our daily lives as AI will have an even bigger environmental impact as it grows.

Dr. Avriel Epps, a computational social scientist and author of the forthcoming “A Kids Book About AI Bias,” puts it bluntly: “Artificial intelligence is making it less likely that society will meet its climate goals to avoid ecological collapse,” they said. With rising global temperatures and unprecedented levels of biodiversity loss, it’s understandable why many people are feeling like AI is the proverbial straw that breaks the camel’s back.

“This is because AI is incredibly resource-intensive and is stressing our power grids, requiring a similar amount of energy as a small Western country. It’s also intensive in its usage of water, precious metals that have to be mined in ways that are environmentally degrading. At a time when we need to be figuring out how we will reduce our emissions and de-grow our economy, AI is a huge contributing factor in us not being able to do that.”

“AI is incredibly resource-intensive and is stressing our power grids, requiring a similar amount of energy as a small Western country.”

– Dr. Avriel Epps, computational social scientist and author

There are three main types of AI in use today: Narrow AI (built for specific tasks like ChatGPT, Google Translate, or Spotify recommendations), limited memory AI (which uses past data to predict future patterns, like predictive text), and machine learning (which analyzes data to make decisions, like in voice assistants). All of these require energy-hungry data centers to function. When Dr. Epps talks about AI’s environmental cost, they’re especially referring to large language models like GPT-4 (which powers ChatGPT) or Gemini (Google’s proprietary model).

Professor Shashi Shekhar, the Institute Director of AI-CLIMATE at the University of Minnesota, says that the rising popularity of chatbots and large language models (LLMs) in recent years has attracted attention to AI, triggering a passionate debate on pros and cons — and the recent growth in energy demand has co-occurred with the rise in the number of data centers and popularity of AI. “This is not sustainable over decades without newer technologies such as resource efficiency and renewables,” he said.

However, many experts like Dr. Epps and Dr. Shekhar agree that it’s not entirely a doomsday scenario. If used correctly and ethically, artificial intelligence can actually be an effective tool in the fight against climate change.

“There are some uses of AI in the climate change mitigation space that are useful — like researchers that are using AI to forecast climate change and natural disasters to try and help keep people safe, or run large-scale simulations of what switching to renewables could look like,” says Dr. Epps. “AI that can help society figure out how to best use our resources to avoid the worst of climate change.”

“If used correctly and ethically, artificial intelligence can actually be an effective tool in the fight against climate change.”

But even the most “efficient” AI model will never be carbon-neutral — just as you can’t shop your way into sustainability, you can’t use AI your way into it either.

As for individual usage, such as turning to ChatGPT for research or productivity — planning a plant-based meal, finding remote work options, or reducing unnecessary purchases — the climate cost is real, but nowhere near the worst offenders. Especially because, according to Dr. Shekhar, AI accounts for such a small share of global emissions when compared to sectors like transportation, energy, and agriculture.

“The last thing I would want to say is that AI is definitely getting us farther away from where we need to be for our climate goals, but it’s still a relatively small piece of the emissions pie. Things like cattle farming and meat-based diets, aviation, and cars are much bigger contributors to climate change still. Your individual usage of ChatGPT matters for the climate, but eating hamburgers does more harm to the planet,” says Dr. Epps.

“According to Dr. Shekhar, AI accounts for such a small share of global emissions when compared to sectors like transportation, energy, and agriculture.”

Where AI’s environmental toll becomes harder to ignore is at the industrial level. These commercial-scale uses — training large models, enterprise-level data mining, or content farms — demand enormous energy and computing power. But as Shekhar points out, many companies are already exploring ways to reduce that impact. “We’ve seen data centers move to cooler regions to lower cooling costs, adopt renewable energy, or shift workloads to areas with cleaner grids.”

Recent growth in data centers is not sustainable over decades without newer technologies such as resource efficiency and renewables. Also, geographic areas with water and energy constraints may have higher impacts. “One approach is to anticipate potential harms and proactively take steps. Some of this has been taking place in data centers for the last two decades due to economic reasons,” says Dr. Shekhar.

“If ChatGPT can be used to help you figure out how to transition to a plant-based diet or how to minimize your air or car travel for work, then we might consider it a net positive on an individual level.”

– Dr. Epps

Of course, he says, more can be done. For example, users may compare the resource efficiency of alternative AI (e.g., chatbot) brands and choose ones with lower environmental impacts based on comparison data from neutral third parties (e.g., Consumer Reports). But beyond that, many are finding ways to use artificial intelligence to further research around best practices in environmental conservation. “If ChatGPT can be used to help you figure out how to transition to a plant-based diet or how to minimize your air or car travel for work, then we might consider it a net positive on an individual level,” Dr. Epps points out.

AI is already being used to reduce emissions from agriculture to navigation. Dr. Shekhar cites eco-routing in apps like Google Maps as one small but meaningful example. “Broadly, AI has the potential to improve resource efficiency across transportation, energy, industry, and agriculture,” he says.

Artificial intelligence is a tool like any other, but it’s a tool being shaped by powerful hands with profit-driven priorities. In the face of the climate crisis, neutrality isn’t an option. The way AI is developed and scaled today reflects decisions that prioritize growth over sustainability. As we’ve seen, it is possible to use AI for good — if companies and researchers are intentional about it.

Sara Jin Li is an essayist, playwright, and filmmaker based in Los Angeles, California. She is also the founder of Heretics Club, a literary salon for creative writers. You can find her on Instagram at @sarajinli.