Artificial intelligence (AI) is rapidly becoming an integral part of modern society, offering solutions to complex challenges and enhancing productivity. However, its environmental impact remains a topic of debate. While many experts believe AI can be a potent tool against climate change, concerns about its carbon footprint persist.
The United Nations Environment Program has praised AI for its potential to enhance our understanding of environmental impacts and climate change. Sasha Luccioni, an expert in AI models for sustainability, emphasized AI’s capability to process vast amounts of data, such as satellite imagery, to monitor climate change. This data-driven approach can help scientists model climate patterns, discern trends, and devise effective mitigation strategies.
AI’s applications extend beyond just monitoring. It can be harnessed to conserve water, combat wildfires, and even streamline recycling processes. “From optimizing electricity grids to tracking biodiversity, AI has diverse applications in the climate change sector,” Luccioni noted.
AI’s Carbon footprint growing concern
However, the environmental cost of AI cannot be ignored. Shaolei Ren, an associate professor at the University of California, Riverside, highlighted the environmental expenses associated with the raw materials required for computer hardware. Training AI models is energy-intensive, and while exact figures remain elusive due to the lack of transparency from AI companies, some estimates are alarming.
A study led by Ren suggests that training GPT-3, the engine behind ChatGPT, might have consumed up to 700,000 liters of freshwater. This water, primarily used for cooling data centers, often evaporates and cannot be reused. Furthermore, researchers from the University of Massachusetts, Amherst, found that training a single AI model could emit over 626,000 pounds of carbon dioxide, equivalent to the emissions from 62.6 gasoline-powered cars driven for a year.
When questioned about their carbon footprint, AI models like Bard and ChatGPT acknowledged the challenges in providing accurate estimates. ChatGPT emphasized that while it doesn’t have a direct carbon footprint, the servers hosting and powering it certainly do. Microsoft, a significant investor in OpenAI, refrained from sharing specific figures but acknowledged the need for a global clean energy supply to power AI technologies.
Designing eco-friendly AI systems
Junhong Chen, a professor at the Pritzker School of Molecular Engineering, stressed the importance of designing AI systems with environmental considerations in mind. Google’s research indicates that water-cooled data centers produce about 10% fewer carbon emissions than their air-cooled counterparts. As data centers account for approximately 2% of the U.S.’s total electricity consumption, choosing eco-friendly cooling methods can make a significant difference.
Ben Townsend, Google’s head of data center sustainability, compared data centers to personal computers, emphasizing their requirements for space, energy, and cooling. He also highlighted Google’s commitment to using reclaimed and nonpotable water resources for their data centers.
Ram Rajagopal, leader of the Stanford Sustainable Systems Lab, pointed out the potential challenges as AI use scales up. The increasing demand could lead to more extensive data centers, resulting in higher power consumption.
AI’s positive environmental impact
Despite concerns, AI’s positive contributions to environmental causes are undeniable. AI models can assist in recycling and reusing water, identifying contaminants, and even suggesting ways to repurpose them. Collaborations between tech giants and other sectors are also yielding promising results. For instance, a joint project between Google, American Airlines, and Breakthrough Energy utilized AI to forecast contrails, which contribute significantly to the aviation sector’s global warming impact.
AI’s potential in battery research, particularly for electric vehicles, is another area of interest. Companies like AMP Robotics and MachineX have developed AI tools to enhance recycling efforts. AMP Robotics claims its technology has helped prevent nearly 1.8 million metric tons of greenhouse gas emissions.
In California, AI is being employed to combat wildfires, with Cal Fire Battalion Chief David Krussow labeling the technology a “game changer.” The National Oceanic and Atmospheric Administration and the United Nations Environment Program are also leveraging AI to improve climate models and monitor greenhouse gas emissions.
Balancing AI’s promise and impact
While AI’s potential in addressing environmental challenges is vast, its environmental cost remains a concern. Kate Brandt, Google’s chief sustainability officer, emphasized the need for a balanced approach. “If we harness AI to address environmental issues, we can positively impact the planet. However, if misused, the consequences could be detrimental,” she said.
As AI continues to evolve and integrate into various sectors, a holistic approach that considers both its potential benefits and environmental costs is crucial.