How AI could save the environment
Advances in artificial intelligence (AI) could be one of the solutions to solving major global environmental crises-from climate change to animal endangerment to disease containment-as projects in each of these areas are already underway.
Business decision-makers working in environmental sustainability are optimistic about the power of AI, according to a 2018 report from Intel: 74% of the 200 professionals in this field surveyed agreed that AI will help solve long-standing environmental challenges. However, several challenges prevent AI from being the answer to all of our environmental problems. Top barriers to doing so included cost (33%) and regulatory approval (17%), the report found.
SEE: Special report: How to implement AI and machine learning (free PDF) (TechRepublic)
Environmental problems typically involve complex processes that scientists do not yet fully understand, and for which we have limited available resources, said Bistra Dilkina, associate director of the USC Center for AI in Society and a member of the Association for Computing Machinery. With advances in machine learning and deep learning, we can now tap the predictive power of AI to make better data-driven models of environmental processes to improve our ability to study current and future trends, including water availability, ecosystems wellbeing, and pollution, she added.
AI can also play a key role in enhancing environmental decision and policy-making work, by bringing an algorithmic approach to that work, Dilkina said.
"Realizing the potential of AI to inform environmental sustainability and social challenges will be key to making progress toward the Sustainable Development Goal in the tight race with irreversible changes in our planet," Dilkina said. "Industries, which stir major technological pushes, can help by understanding the computational needs for informing sustainability issues, identifying gaps and opportunities in current AI methods, and prioritizing AI work that leads to broader impacts."
Environmental AI applications
AI has already been applied to several environmental problems, said Brandon Purcell, principal analyst at Forrester.
For example, WildTrack uses a computer vision solution developed by SAS called Footprint Identification Technology to monitor endangered species non-invasively. The tool analyzes images of footprints of cheetahs, rhinos, and other endangered species to identify them, track them, and determine what threatens them, Purcell said.
SEE: Special report: Managing AI and ML in the enterprise (free PDF) (TechRepublic)
Google has also used its own AI expertise to improve its energy efficiency as a company-leveraging DeepMind's machine learning capabilities, it reduced the amount of energy needed to cool its data centers by 40%.
"AI is most helpful when the possible solution to a problem resides in large, highly dimensional datasets," Pucell said. "If you think about climate data, there's a wealth of traditional structured data about temperature, sea levels, emissions levels, etc. But there's also a lot of unstructured data in the form of images, video, audio, and text. When it comes to analyzing massive amounts of unstructured data, deep learning is really the only game in town."
At USC, Dilkina's research group has used AI to develop optimization methods for wildlife conservation planning-an area where highly limited budgets need to be allocated to protect the most ecologically effective land, she said.
Her team has also used machine learning and game theory to help protected areas fight the poaching of endangered animals, including elephants and rhinos. "By leveraging historic data, we can create behavioral models of poachers and use them to optimize patrolling strategies to defend against them," Dilkina said.
The explosion of Internet of Things (IoT) data, including sensors ranging from cameras in a field to satellites in space, has helped scientists use AI techniques to extract more data, Dilkina said.
For example, advanced AI and vision techniques help biologists detect animals in pictures from camera traps to study their distributions. And deep learning has created models that use satellite images to recognize different forms of land cover worldwide, enabling urbanization and forest and water change detections, and informing agriculture, climate studies, and health studies. It has even been successfully used to detect levels of poverty from space, Dilkina said.
AI applications also track mosquito populations to anticipate or prevent the spread of disease, as well as weather changes, to warn populations about upcoming storms, said Erick Brethenoux, a research director at Gartner.
In California, some vineyards use AI to determine if vines are getting enough water and sunlight. "It's like crop personalization," Brethenoux said. These techniques also prevent farmers from using too much water unnecessarily, he added.
How businesses can leverage AI to benefit themselves, and the environment
AI techniques offer an advantage in environmental work, because they are able to process large amounts of data very quickly, and draw conclusions from that information that humans may not have the capability to see otherwise, Brethenoux said.
"Sometimes there's a very small change that ends up being a big problem, but we miss it because we are not that detail-oriented when it comes to very large amount of data," Brethenoux said. "We focus on the big groups rather than on the details. It's not just particular to environmental situations, or applications, or solutions. But when we go to do that, it becomes very important."
The ability to solve major world issues with AI depends on our ability to gather large data sources on these problems, Purcell said. And many valuable data sources on environmental issues currently reside within businesses, he added.
"To really get real value from AI, businesses will need to cooperate to share this crucial data," Purcell said. "At the end of the day, good environmental stewardship is good business, and good data builds good AI."
Organizations not directly involved in environmental causes can still often benefit from using AI to track these factors, Brethenoux said. For example, a US insurance firm has models to track hail, which damages cars and leads to expensive repairs. When hail is predicted in the forecast, the company can text its clients and ask them to park their car in a garage if possible, or avoid the area.
"It depends on the use case-it's about trying to find the real value that can be derived very quickly from the technology," Brethenoux said.
Despite the recent hype, many AI techniques have been around for decades, and have already provided a number of benefits for the environment and other sectors, Brethenoux said.
"When we start to move away from the hype and realize the real, tangible value that these techniques can provide by being applied to business problems or environmental problems, then we can start to see the real benefits," Brethenoux said. "Expectations and the practical implementation of those techniques is critical."