At a time when the world is focused on accelerating zero-carbon transition, applications in artificial intelligence (AI) and data science have been developed with an eye on salvaging the environment. AI is applying the power of machine learning to find patterns in data across sectors, spot trends and monitor Earth’s resources better. However, use of AI to fight climate change comes at a cost, say experts.

Researchers believe that before using AI to build climate solutions, it is important to understand its growing carbon footprint. AI can be used to strengthen climate predictions, make smarter decisions for decarbonising industries and allocating renewable energy.

Also read: How is AI slowly eating up white-collar jobs?

However, AI’s relevance as a climate change fighting tool comes amid increasing ethical concerns linked to machine learning applications, which use data to make predictions and decisions. Data usage may give rise to public surveillance, privacy issues, bias and intentional misuse, resulting in discrimination and inequality.

AI, a carbon emitter

AI itself is a significant emitter of carbon. In 2019, researchers at the University of Massachusetts Amherst examined several natural language processing training models available online to understand the energy cost required to train them.

The study estimated that the carbon footprint of training a single big language model would emit 300,000 kg of carbon dioxide, a Nature report said. The total emission is equal to almost 125 round-trip flights between New York and Beijing.

Also read: Metaverse: A blessing or curse for environment?

With the rise in demand for AI, there is a need for more processing power associated with larger AI models. In 2020, digital technologies were responsible for 1.8 percent and 6.3 percent of global emissions, Indian Express reported.

Inequality

Only a few developed economies have material advantages, strength in research and development, skilled workforce and the wealth to invest in AI. According to the Indian Express report, East Asia and North America alone are responsible for three-fourths of the world’s private investment in AI, patents and publications. Climate change could deepen this inequity among nations, the report said.

Finding a solution

Experts believe that environmental sustainability should be one of the major principles guiding responsible development and application of AI.

“It is important to note that AI is not only just a tool but a resource demander…

Various steps can be taken to ensure cleaner AI practices. Green AI certifications could be given for industry processes that promote green AI development.

Also read: Blockchain’s sustainability problem – Not so clean, not so green

It is important to set up a framework and guidelines for organisations and companies to support environmentally-friendly AI practices, Gabriela Prata Dias, head of the Copenhagen Centre on Energy Efficiency, and Xiao Wang, programme officer, told Nature.

Also, governments should work on setting up regulatory frameworks and legislations that would legally address transparency and sustainability in AI development.

In the context of AI’s climate costs, governments should reconsider and assess their technology-based growth priorities. Deepika Sandeep, AI scientist at Bengaluru-based clean energy generation company Bharat Light & Power, told Nature that a judicious approach was required for use of deep learning technologies. “Not every problem demands a machine learning-based solution,” Sandeep said.

It’s a positive start

Adopting the ‘Recommendation on the Ethics of Artificial Intelligence’ in November 2021, the UNESCO urged global players to bring down the environmental impact of AI systems, “including, but not limited to its carbon footprint.”

Also read: Explained: Big tech betting big on green energy to reduce carbon footprint

Tech giants Microsoft, Amazon, Google parent Alphabet and Meta have announced net-zero policies and initiatives. Although these initiatives are a positive move, it is only a scratch on the surface.

(Edited by : Shoma Bhattacharjee)