The impact of artificial intelligence on the climate goes beyond its emissions; it also relates to how it is used.
The Impact of Artificial Intelligence on the Climate
Artificial intelligence is more than just entertainment programs, as effective programs trained with machine learning have become a fundamental part of various applications such as smartphone navigation and social media algorithms. However, with the increasing popularity of artificial intelligence, more researchers and experts are noticing the environmental cost of it. Training and operating an AI system requires a significant amount of computational power and electricity, and the resulting carbon dioxide emissions are one way that AI impacts the climate. But its environmental impact goes far beyond its carbon footprint.
The Impact of Artificial Intelligence on the Environment
It is difficult to calculate the precise impact that artificial intelligence will have on the climate crisis, even if experts focus solely on the amount of greenhouse gases it emits. This is because different types of artificial intelligence, such as a machine learning model that discovers trends in search data, a vision program that helps self-driving cars avoid obstacles, or a large language model that enables chatbots to converse, require varying amounts of computational power for training and operation. For example, when OpenAI trained its large language model GPT-3, this effort produced roughly 500 tons of carbon dioxide. However, simpler models produce negligible emissions. Moreover, it is complicated by the lack of transparency from many AI companies, making it difficult to understand the impact of their models when examined solely through the lens of emissions.
Harmful Environmental Applications of Artificial Intelligence
The application of harmful environmental artificial intelligence is not limited to the fossil fuel industry. There are examples of this across all sectors, such as forestry, land management, and agriculture. This can also be seen in how AI is used in automated advertising. When a specific ad weirdly appears on the news feed of Instagram or Facebook, the ad algorithms are the magicians behind the curtain. This practice promotes overall consumer behavior in society; for instance, with fast fashion ads, targeted advertising drives a continuous flow of cheap, mass-produced clothing to consumers, who purchase fashion only to replace it as soon as a new trend emerges. This leads to increased demand from fast fashion companies, which are estimated to produce up to eight percent of global emissions. The fast fashion industry generates more emissions from shipping and contributes to the accumulation of discarded clothing in landfills. Meta, the parent company of Instagram and Facebook, did not respond to Scientific American’s request for comment.
Beneficial Environmental Applications of Artificial Intelligence
On the other hand, there are AI applications that can help address climate change and other environmental issues, such as the destruction caused by heat-driven hurricanes. One such application is the xView2 program, which combines machine learning models and computer vision with satellite imagery to identify damaged buildings in natural disasters. The program was launched by the Defense Innovation Unit, an organization within the U.S. Department of Defense. Its models can assess damaged infrastructure, reducing risk and providing time for first-response teams who would have to conduct those assessments themselves. It can also help identify locations where rescue teams should direct their efforts more quickly.
Other AI techniques can be applied directly to combat climate change by monitoring emissions. Gavin McCormick, CEO of WattTime, which monitors electricity-related emissions, says that in most of the world, and for most climate change emissions, they are very opaque. WattTime is a co-founder of the nonprofit Climate TRACE, which combines computer vision and machine learning to identify emissions from global pollution sources. First, scientists identify emissions from monitored facilities. Then they use satellite images to identify visual signatures of activities causing emissions, such as steam plumes from a factory. Afterward, engineers train algorithms on that data to teach the programs to estimate emissions based solely on visual inputs. The resulting figures can then help companies identify ways to reduce their emissions impact and can provide information to policymakers and aid in holding polluters accountable.
The improvement of artificial intelligence efficiency in solving environmental problems, such as reducing emissions, can be a valuable tool in combating climate change, if the AI industry can reduce its negative impact on the climate. David Rolnick, a computer scientist at McGill University, states that from a policy perspective, there is a role in shaping AI policy in a way that considers all aspects of its impact on the climate. This means looking at its applications as well as its emissions and other production costs, such as water usage.
Additionally, Jesse Dodge adds that people with expertise in artificial intelligence, especially those with influence in tech companies, should establish ethical principles to restrict the use of technology. The goal should be to avoid environmental harm and instead help reduce it. “It should be part of the value system,” he says.
Jud Cohen is a freelance science journalist based in Oregon covering stories about the environment, climate change, and ecology.
Source: https://www.scientificamerican.com/article/ais-climate-impact-goes-beyond-its-emissions/
Leave a Reply