By Dr Leslie Kanthan, Co-founder and CEO of TurinTech

AI and automation are everywhere. Historically, most industries have depended heavily on technological advancements to improve their offerings and stay ahead of competitors. As a result, the development of the global artificial intelligence industry has been significantly boosted by the increased usage of digital technology and the internet.

The significant investments in R&D from tech giants is continuously advancing technology across a range of industries. In the upcoming years, the global artificial intelligence market is expected to be driven by the escalating demand for AI across a number of end-use industries, including automotive, healthcare, finance, manufacturing, and retail.

However, one particular issue that is often overlooked in this industry is the sustainability of its processes.

AI itself has an extremely high carbon footprint. If no attempt is made to reduce this, there will be drastic consequences. This year, we have seen the extent of climate change in action with temperatures exceeding 40 degrees in the UK. 

Every single person, organisation, and institution must do more to protect our world, and this includes the AI industry. Organisations need to begin working together to form greener forms of AI. Technological advancements are not set to slow down in 2023, but neither is the earth’s rising carbon footprint level. Currently, AI is effectively improving efficiency within traditional systems, but can it be doing more? 

The pros of AI are too good to lose, but the cons are growing too difficult to ignore. Therefore, how can AI be used to mitigate one of the problems it is currently contributing to. 

Environmental effect of AI

Let’s look at the numbers. The amount of computing capacity required for the largest AI training runs has increased by 300,000 times since 2012. 

Some models could have consumed as much as 67,000 gallons of gasoline as they were being trained. There are also some calculations that show that even smaller models can consume on average over 9,000 kilowatt-hours of electricity per day – which is nearly as much as the yearly average household electricity consumption in the UK. A University of Massachusetts Amherst study estimates that training a single Natural Language Processing model can generate nearly five times the lifetime emissions of the average English car.

So, how is this cause? The intensive use of Graphic Processing Units to train machine learning models contributes to rapid CO2 emissions. These emissions cause many problems such as global average temperature rise, sea-level rise, and air pollution, to name a few. With more companies deploying AI at scale, and the ongoing energy crisis, this is an issue that needs to be addressed and the scrutiny on this issue will only continue to grow as the climate crisis deepens. 

What regulation could be put in place? 

There have been many formal and informal initiatives to hold countries and big private companies accountable for their carbon footprint. However, most regional government policies, such as the UK’s AI Rulebook and the EU’s AI Act, fail to give notice to the profound environmental impact of AI. Even the recent ‘AI in action’ released by the UK government failed to provide any mention of the environmental impact of AI. 

Regulations such as monitoring AI-related emissions, choosing more energy-efficient models and using cleaner energy for building AI models could be implemented to combat the ongoing issue. 

This being said, specific regulation targeting AI would not be sufficient alone. A Harvard University research paper found that information and computing technology is expected to account for as much as 20% of global energy demand by 2030; with hardware responsible for a majority of that footprint. Regulation would have to cover chips and other technologies underlying AI to address environmental impact concerns.

AI leveraged to increase AI’s efficiency

As we look forward towards 2023, we will begin seeing AI be used in more ways to help protect our future. Some companies have already started to optimise their existing AI models by using AI-powered base code optimisation techniques to address energy use and carbon emission concerns before deploying a machine learning model. 

AI can help combat its own carbon footprint by speeding up model development and AI code efficiency. Using AI that is more efficient to improve existing AI can lead to better sustainability in our economy.

There are a few ways AI can help tackle climate change in the future: 

Accurate Demand Forecasting

By anticipating demand (e.g. for energy, retail goods) a distributor can reduce overheads in their system and increase efficiency. Businesses may minimise errors by 20% to 50% if they implement AI-driven forecasting in their supply chains, according to McKinsey. This can also be supported by optimised logistics which can help find the best routes to transport goods and passengers. 

Improve efficiency 

An example of improving efficiency in the STEM field is finding drug candidates by using fewer computational resources than traditional methods. There is even the possibility of comparing the efficacy of competing decarbonisation strategies. Experiments can be designed to understand the impact of each strategy containing specific proportions of Carbon Capture and Storage (CCS), geothermal, and renewable technologies.

By measuring the collective impact of CCS and heat-pump deployment effects on the whole energy system, energy policy strategies can be designed for a greener tomorrow.

Monitoring environmental degradation

AI can also help us to monitor the Earth for signs of environmental degradation. It can predict the failure of a plant or a windmill or measure the degradation of solar panel conversion rate in real-time.

AI allows us to process satellite images and identify various kinds of pollution including oil spills and rubbish in the ocean, as well as wildfires and other kinds of habitat loss. If paired with an effective and timely response, such Earth observations can help us to protect the environment.  

The bottom line

The issue continues to exist, and the scrutiny on it will only grow in the future. AI is one of many causes of climate change, but it may be one of the few that could also be used to combat the crisis. The next step forward in creating a greener future is using AI to find innovative ways to deal with the current problem.