AI and Its Environmental Impact: A Conservative Perspective
Marissa Loewen began utilizing artificial intelligence in 2014 as a tool for project management. Living with autism and ADHD, she found that AI significantly aided her in organizing thoughts. “We’re mindful of its environmental impact,” she noted.
Today, AI is integrated into our daily lives, from smartphones to search engines and emails. However, every interaction with AI consumes energy, often produced from fossil fuels, which contributes to greenhouse gas emissions and climate change. As a result, many find it increasingly challenging to navigate life without AI support.
The Environmental Costs of AI
AI primarily operates through data centers that handle queries and store information. As AI technology becomes more widespread, the demand for these data centers is escalating, leading to reliability issues for communities nearby. Noman Bashir, a fellow focused on computing and climate impact at MIT, pointed out that most new data centers are reliant on fossil fuels due to the current pace of development.
These data hubs also generate considerable heat and require vast amounts of water for cooling. The larger centers can consume up to 5 million gallons of water each day, equivalent to the needs of a town with 50,000 residents.
For many users, the environmental ramifications are not immediately apparent. Sasha Luccioni, who specializes in AI and climate, highlighted that creating a high-definition image can use as much energy as charging half a smartphone, which many people find hard to believe due to the lack of visible impact.
Jon Ippolito, a professor at the University of Maine, noted that while tech companies strive to enhance efficiency, this doesn’t necessarily lessen AI’s environmental footprint. He referenced the Jevons Paradox, suggesting that as resources become cheaper, we tend to use them more without reducing our overall consumption.
Measuring AI’s Environmental Footprint
Determining AI’s exact contribution to global warming is complex, influenced by factors like outside temperatures, energy grid cleanliness, and the complexity of the AI task. However, estimates can be made. Ippolito developed an app that compares the environmental costs of various digital tasks, revealing that a straightforward AI prompt can consume 23 times more energy than a traditional Google search without AI features.
For more complex requests, such as calculating the number of gummy bears that could fit in the Pacific Ocean, the energy consumption can spike to 210 times that of a simple search. Shockingly, a brief video can use 15,000 times more energy than a basic text inquiry. Yet, it’s essential to recognize that our digital footprints were already substantial before AI entered the picture. For instance, an hour of Netflix consumes more energy than a complex AI prompt.
Reducing Our Tech Impact
Ippolito advocates for minimizing AI usage when possible. For instance, opting for human-captured images over AI-generated ones and limiting unnecessary prompts can save energy. Loewen has adopted similar strategies, condensing her queries into one instead of multiple. She even developed her own AI solution that doesn’t rely on large data centers, effectively reducing energy consumption similar to watching a DVD rather than streaming.
Luccioni chooses Ecosia, a search engine that prioritizes efficiency and uses profits to plant trees for mitigating the impact of searches, while still allowing the user to disable AI functionalities.
ChatGPT’s temporary chat feature ensures queries are deleted after a few weeks, thus conserving storage space. Importantly, AI accounts for just a fraction of data centers’ energy consumption; around 85% usually comes from data gathered on social platforms and cryptocurrency.
To address this, Ippolito recommends leveraging screen time limits on devices to curtail social media use. Less scrolling means reduced personal data collection, lower energy consumption, and fewer emissions. “If you can eliminate a data center from the equation, that’s a win,” he concluded.
In light of these insights, it’s clear that while AI provides many benefits, we must balance its usage with an awareness of its environmental implications.
