Climate change: can AI contribute to a more sustainable future?

Artificial intelligence (AI) promises to revolutionize many areas of daily life, taking data and turning it into actionable insights that drive positive results and benefits. AI’s power is now being brought to bear on arguably the biggest challenge of our times: climate change. But it isn’t without problems.

One of the unexpected byproducts of the COVID-19 pandemic was a positive impact on the environment. Lockdowns, flights suspended and people worldwide not commuting to work all contributed to perceived bluer skies, metaphorically and literally.

However, in 2020, CO2 emissions were still 80% of the 2019 total, even with a global economy in near shutdown at times. If we don’t act decisively now, it’s estimated that the economic damage caused by climate change over the next two decades could be as detrimental as having a COVID-type pandemic every ten years. Can digital technology play a part in the solution?

How can AI help?

AI and machine learning (ML) have positively impacted the commercial world in recent years. According to research by PwC, 54% of executives say that deploying AI solutions has increased productivity in their businesses.

On the issue of climate change, AI and ML can help boost productivity without increasing emissions. Take the example of a factory, where emissions need to be reduced. Smart factories can use connected IoT devices to track parameters like location, temperature and flow rates. They can then use AI and ML to optimize the flow of materials and resources throughout a product lifecycle.

This marriage of AI and ML with IoT solutions in factories creates a seamless, connected network of machines and people in a production environment, which can collect and analyze data and act on it in real time. This can allow them to reduce energy consumption and carbon emissions in manufacturing processes. In fact, monitoring and optimizing production processes with AI and ML can lead to a 20% reduction in annual energy intensity by manufacturing companies.

Energy consumption in buildings is another area ripe for optimization. Energy consumed by buildings comprises around a quarter of global energy-related CO2 emissions, and yet it’s an area that should be among the easiest to address.

By adding smart sensors to a building to monitor air temperature, water temperature, lighting and energy use, and analyzing it with AI tools, you can reduce energy usage by up to 25% in a single building. This can be achieved by automatically controlling temperature or lighting as needed in particular parts of the building. That’s something that can scale up rapidly, and using AI to monitor and address energy usage across whole cities could have an even greater impact.

Focus on data

Data is at the center of how and why AI and ML can help fight climate change. With data analysis, we gain insights and knowledge. This helps highlight areas that need the most urgent attention and enables us to make better-informed decisions. It could be data about deforestation rates that helps policy makers and law enforcement take action or data on sea surface temperatures that aids future weather prediction over land regions.

AI can empower us to make better electricity systems, for example. Electricity networks are full of useful data about consumption, peak usage days and times, and more. ML can take this data and make forecast models around electricity generation and demand.

AI and ML can also enable predictive models in renewable energies like wind power. Engineers regularly face challenges in predicting weather changes and, in turn, calculating supply and demand ratios. AI and ML tools can use algorithms built using data from historical weather forecasts and data from wind turbines themselves to make accurate predictions, often up to 36 hours before the actual energy generation. This lets operators schedule and optimize energy input into the power grid a whole day in advance, reducing waste.

Google UK’s DeepMind project has already made a case for this type of initiative, using AI to predict the energy output of wind farms, with this type of predictive model. It reports positive results, with AI and ML having boosted the value of wind energy generated on its farms by around one-fifth.

Detecting environmental change

AI can also play an important role in detecting environmental issues, such as deforestation or water quality issues. Using satellite imagery and AI, researchers can more accurately identify where deforestation is taking place and put protective measures in place around these natural carbon sinks.

The organization Rainforest Connection utilizes strategically-placed cellphones in Amazon basin countries to build acoustic monitoring systems that notify rangers when the sound of a chainsaw is detected. The system acts as an early warning against illegal deforestation and uses AI to build better predictive models about where the deforestation might move to next.

Some of the most significant impacts of climate change in years to come will be caused by extreme weather events and systems, such as changes in cloud cover and ice sheet dynamics. Traditionally, meteorologists use masses of data and huge computing power to make weather forecasts based on relatively basic predictive models. The data is gathered from sources like deep-space satellites, weather balloons and radar systems, sources that are prone to faults and inaccuracies. The algorithms AI utilizes can fill any gaps in meteorologists’ data collections using previous data sets and the chains of information and enable more accurate forecasts.

Climate change: can AI contribute to a more sustainable future?

Concerns and challenges remain

However, for all the possibilities, concerns remain. Concerns about privacy and ethics in AI remain, and the massive processing requirements for AI and ML currently contribute too much to emissions. In 2019, researchers at the University of Massachusetts Amherst analyzed various natural language processing (NLP) training models to establish the energy cost in kilowatts required to train them.

They converted the energy consumption to ballpark carbon emissions and electricity costs and found that the carbon footprint of training a single big language model equated to around 300,000 kg of CO2 emissions. In context, that is around 125 return flights between New York and Beijing. According to Virginia Dignum, Professor in Social and Ethical AI at Sweden’s Umea University, “AI is both an enabler and, potentially, a destroyer of the climate fight.” These concerns need to be tackled if AI and ML are to fulfill their potential as powerful tools in the fight against climate change.

Orange is committed to being a net-zero company by 2040, and sustainability is a central tenet of our Engage 2025 strategic plan.