State officials faced a dilemma when India’s government, states, and union territories began the 24 x 7 Power for All program, an endeavor to give national access to energy. They had to purchase electricity power blocks in advance, which required them to precisely forecast demand. Calculations on paper were insufficient.
Artificial intelligence models, thankfully, came to the rescue. Models aid in the prediction of electricity usage, allowing budgets to be saved and resources to be saved.
Machine learning, a type of AI, is based on the idea that algorithms can be trained to detect patterns and images. The algorithms can then utilize this information to find comparable events in enormous amounts of data.
Shared historical weather data from IBM’s The Weather Company, for example, laid some of the foundation for AI models in projecting electricity usage in India’s states of Uttar Pradesh and Bihar. Electricity providers might proactively estimate demand and states could buy allotments based on past weather patterns preceding electricity consumption spikes and falls, among other things.
This weather information is part of IBM’s Environmental Intelligence Suite, which is also used in a variety of other ESG projects. According to Rishi Vaish, CTO and vice president of IBM AI applications, companies utilize the AI-powered suite to predict potential implications of climate change and weather across their organizations.
“Think of it as a digital doppelganger of Planet Earth,” Vaish said. To give business insights, the suite layers petabytes of satellite data from the Weather Company with time and geolocation data. Helping utility companies determine where to spend their vegetation control funds based on plant growth trends is one example in action.
Artificial Intelligence (AI) in Environmentally-Friendly Programs.
Capgemini, a consulting business, is working on machine learning models for an image identification project in the Mojave Desert that will distinguish between man-made and natural paths. The Mojave Desert, which covers over 47,000 square miles, is home to a diverse range of fauna. Recreational sports such as all-terrain vehicle and off-road dirt bike riding, which can be harmful to the environment, are also popular in the desert. Capgemini conducted the initial phase of the project pro gratis for The Nature Conservancy, constructing AI models that use satellite photos to study the desert environment over time. The project’s ultimate goal is to determine the impact of humans on endangered species and to recommend recreational alternatives.
Environmentally conscious AI projects can also be operational enhancements. Organizations can eliminate waste through reducing inefficiencies in day-to-day operations. For example, IBM’s Maximo employs AI to predict asset maintenance so that they last longer.
AI Data Challenges are Addressed by ESG Initiatives
ESG projects that use AI will run into the same issues that AI deployments in other domains do. “The requirement for solid data, being able to train appropriate models, and making the outcomes explainable to individuals driving the process,” according to Vaish.
As Capgemini discovered, decades-old satellite imagery lacks sufficient resolution for precise ground-level analysis and identification (photos are shot from too far above the earth). It’s possible that patterns will go unnoticed.
“Photos are only a snapshot in time, so they don’t always convey a lot,” said Dan Simion, Capgemini’s vice president of AI and analytics.
Capgemini supplemented their current data sets with generated data. “We used AI to produce synthetic data,” Simion continued, “so we could still train the models if the image quality was poor or if we didn’t have enough instances of a certain kind of image, like [photos of animal] nests.”
Getting user buy-in for ESG efforts that leverage AI can also be difficult. “It’s vital to gain business users’ trust, discover and remove any bias or drift in the data, and operationalize the AI model development-to-deployment lifecycle,” Vaish said. “This is especially true in the areas of asset and supply chain management and environmental intelligence, as each has its own set of data (some from the public domain, some from within an organization, and some from across organizations), as well as its own set of system and application integrations. For an AI solution to be effective and consumable, all of these issues must be resolved.”
Even if these issues are addressed, AI alone will not be enough to propel an ESG program forward. “Organizational commitment to creating ESG goals, investing in and enabling teams to achieve these goals, tracking progress, removing impediments, and evaluating and accounting for success are all required,” Vaish said.
