How Microsoft & Google are Using AI to Fight Climate Change

The technology sector is turning AI towards environmental challenges at a time when the energy demands of machine learning (ML) systems are drawing increased scrutiny from climate researchers and policymakers.

Microsoft and Google are now deploying AI models across programmes that target emissions reduction, weather forecasting and waste management. 

The initiatives are focused on operational systems handling real-world environmental data, though questions remain about whether the technology’s climate benefits outweigh its substantial energy costs.

Google’s CircularNet is one of the more tangible examples. The open-source AI model identifies materials in waste streams to improve sorting accuracy in recycling facilities.

Kate Brandt, Chief Sustainability Officer (CSO) at Google, says on LinkedIn that doubling the global recycling rate “could save more than 10 gigatons of carbon over the next 30 years”.

A start-up using the model reports detection accuracy above 90%, which translates to measurable improvements in recycling operations. 

Kate says this “has improved the purity of their recyclables by approximately 60%, leading to a 6x increase in material value and diverting over 50,000 metric tonnes of waste from landfills every month”.

She adds that this demonstrates “the opportunity of AI in action: using technology to create efficiency, reduce waste and unlock economic growth”.

Why Microsoft is funding wildfire prediction through Vibrant Planet

Google has also deployed its NeuralGCM weather prediction system to help Indian farmers forecast monsoon patterns for crop planning. 

The company’s track record with AI applications includes an earlier project where its DeepMind division cut data centre cooling energy by 40 per cent through algorithmic optimisation.

Microsoft is taking a different route through its Climate Innovation Fund, which backs companies developing AI for environmental monitoring. 

Melanie Nakagawa, CSO at Microsoft, says the fund supports Vibrant Planet, a company analysing vast datasets to predict and manage wildfire threats.

Melanie says “Vibrant Planet’s models integrate observational data, global and regional climate models and ecosystem models, to provide more accurate predictions and effective interventions for wildfires”.

Addressing energy and security concerns 

The technology sector’s environmental credentials face persistent questions over AI’s energy intensity. 

Training and running large ML models requires substantial computational power, raising the possibility that the electricity consumption could negate climate benefits if not carefully managed.

A United Nations Framework Convention on Climate Change report identifies AI as a tool for reducing greenhouse gas emissions through optimisation of energy grids, industrial processes and transport networks. 

The technology can handle datasets and simulations that would overwhelm human analysts, but only by consuming power at scale.

Security concerns add another layer of complexity. 

Environmental infrastructure controlled by ML systems could become vulnerable to cyber attacks if security measures fail to keep pace with deployment. 

The distribution of AI capabilities between wealthy and developing nations presents a parallel challenge, with unequal access potentially widening economic gaps.

The Coalition for Sustainable AI has formed to address both the environmental footprint of ML and its applications for climate solutions. 

Melanie says Microsoft is working to “accelerate the transition to net zero” through partnerships with companies developing AI solutions for environmental monitoring and prediction.