Artificial intelligence has often been mentioned in the same breath as climate change over the past few months. After all, why settle for one existential threat to humanity when you can have two?
The rapid advance of generative AI in particular has caused widespread consternation since the end of 2022. Several high-profile figures have argued that the technology is a more pressing risk than climate change because it will be far harder to put the AI genie back in the bottle than it will be to decarbonise the world.
But is it necessarily fair or helpful to pit the two against each other in such a way? After all, the disturbing forecast that global warming will surpass the hopeful 1.5ºC ceiling targeted by the UN in 2018 was actually the product of AI modelling. If neural networks can serve as an effective early-warning system, surely they have other useful applications in the effort to solve the climate crisis.
Could AI supercharge emissions tracking?
At the World Economic Forum’s meeting at Davos in January, Thomas Siebel, chair and CEO of software firm C3.ai, suggested that AI’s ability to “ingest huge amounts of data” and “pull signal from noise” would make it an important tool in tackling climate change. He suggested that it could be used to more accurately assess firms’ progress towards ESG targets such as emission reductions, for instance.
The insights from AI can help in gauging the efficacy of emission reduction tactics
Siebel is no lone voice. Professor Somdip Dey, an embedded AI scientist and the founder and CEO of Nosh Technologies, says: “There is a growing body of evidence suggesting that meeting ESG targets can have a positive impact on climate change.”
Of course, the sheer volume and complexity of ESG data generated are likely to be daunting even for the best-resourced companies. But this is where AI can help to reduce the burden and ensure that all the correct metrics are tracked, so firms don’t have to work towards their net-zero goals in the dark.
AI can be used to “recognise trends in emissions over time. The insights from that can help in gauging the efficacy of reduction tactics,” Dey says. “And it can automate data acquisition, analysis and reporting. This frees up human resources, allowing more expertise to be channelled into developing and executing reduction strategies.”
Designing more efficient renewable energy sources
AI has already been applied in the generation of renewable electricity. Danish firm Vestas Wind Systems uses the technology to make its wind farms more efficient by adjusting individual turbines so that the air turbulence their rotations cause doesn’t disrupt the intake of turbines downwind of them.
Working with tech partners Microsoft and minds.ai, Vestas applied reinforcement learning to the challenge. This technique is a type of machine learning in which the systems teach themselves a task by learning from environmental changes in real time, gaming out different scenarios and receiving rewards when desired outcomes are achieved. The system ran simulations in which it responded to a whole range of wind conditions and repositioned upwind turbines automatically to optimise the whole farm’s efficiency.
Is AI too energy-hungry to help with climate change?
Artificial intelligence’s potential as a weapon in the battle with global warming and climate change is tempered by a key factor, especially as we enter the GPT-4 era of generative AI: the technology needs vast amounts of energy to work properly.
Can AI save the rainforest?
AI has also been used in the global fight to prevent the conversion of ecologically important carbon sinks into agricultural land.
“We run several machine-learning models to produce Global Forest Watch, an open-source web application,” says Evan Tachovsky, global director of the World Resources Institute Data Lab. “Some models are trained on optical imagery and others use radar imagery, which can help us to see through clouds.”
They can detect tracts of forest that are being cleared by picking out new agricultural plantations based on their colour, size, shape and pattern. “Our systems give us a near-real-time view, with local precision at global scale, of where deforestation is happening,” Tachovsky explains. “We can then issue alerts and serve out data to the relevant audiences.”
What about AI’s hidden bias problems?
If these use cases are anything to go by, AI has plenty of further applications in the struggle to limit global warming and climate change. But bias is an ever-present challenge – a factor that must always be monitored if the technology is to prove truly effective. A whole new discipline, known as responsible AI, is emerging as a result.
“This is where machine learning is limited,” says Dr Kasia Tokarska, a climate data scientist who specialises in applying AI. “A model of the climate system that obeys the conservation of carbon, energy and water can be trusted more than a purely black-box approach where you feed in data and get some results back.”
Tokarska suggests that AI users must always be wary of feeding their systems with data derived only from observed events. This, she warns, can lead to hallucinations, the term applied to wildly inaccurate AI outputs. They should instead ensure that new events – new government policies concerning the environment, for instance – are also included in AI’s data diet.
