Green AI: How can AI solve sustainability challenges?

Deloitte has called for businesses and governments to commit to ‘Digital with Purpose’, maximising the positive impact that AI and other digital technologies can create. Embracing ‘Green AI’ is certainly one way that businesses can support a more sustainable future.

Now is a particularly opportune time to drive towards this goal. As the world moves towards a COVID-19 post-pandemic recovery, the UN has called on governments to heed the “unprecedented wake-up call” and “build back better” by creating more sustainable, resilient and inclusive societies.

So how exactly can AI help? There are two approaches to Green AI – using AI to solve sustainability challenges and using AI in a more sustainable way.

How can Green AI solve sustainability challenges?

Delivering societal and environmental well-being through AI are key strategic considerations of the European Commission, who acknowledge that “AI systems promise to help [tackle] the most pressing concerns, including climate change and environmental degradation”.

The UK Government has also recognised that Green AI can help to address the UK’s Grand Challenges, which are four transformative global trends set out in the UK Industrial Strategy, which include issues related to climate change. Efforts include £200 million of funding towards 1,000 new PhD places over the next 5 years, for studying AI which could make industries more sustainable. Following the publication of the AI Sector Deal in April 2018, partnerships between the UK Government Office for AI, the Open Data Institute and Innovate UK have supported many sustainable initiatives, such as global food waste reduction efforts. These partnerships have also worked on tackling the illegal wildlife trade, using algorithms to classify images of illegal animal products. We expect to see increased government interest in applying Green AI, particularly in Smart Cities such as around energy network management.

Recently, a series of climate change-related workshops were run by the Climate Change AI working group at the (virtually-hosted) International Conference on Learning Representations (ICLR) 2020, one of the premier global machine learning conferences. Themes of AI application areas included: energy modelling for infrastructure optimisation and urban planning; utilising diverse data sources for environmental monitoring and targeted sustainability (bioacoustics, Internet of Things (IoT) water sensors, soil analysis, satellite imagery); and the acceleration of climate science (physics emulators, climate forecasting, materials science).

There are many other ways that AI can be used to support sustainability, but this will rely on partnerships between academia, industries and governments to create action and achieve lasting impact.

The circular economy

For the Climate Change & Environment Studio in Deloitte Ventures, the circular economy is an area of increasing importance. This concept has three key principles:

1. Keeping products and materials in use – integrating AI, IoT and geographic information system (GIS) technologies to track, aggregate and close the loop on secondary materials (such as recycled plastic) as it moves through the supply chain - preventing resource loss and environmental damage.

2. Regenerating natural systems – building upon precision agriculture examples from ICLR 2020, improved produce quality and yield can go hand-in-hand with improved stewardship of ecosystem services, as well as climate change mitigation.

3. Designing out waste and pollution – Green AI can streamline product development and create responsive and sustainable supply chains, accelerating business decarbonisation and resilience to environmental and other crises.

Using AI in a more sustainable way

Although it can sometimes feel like it, AI is not a ‘silver bullet’. As with any technology, AI must be used responsibly.

There has been a trend in creating larger AI models to give better performance on tasks, especially in Natural Language Processing (NLP). Larger models (meaning those with a larger set of tuneable parameters) require more data and computation power to train and run, creating more emissions, cost, and technical barriers-to-entry for developers. Developing and deploying ‘greener’ AI should be a consideration for alleviating these issues.

More positively, the AI ecosystem is strongly integrated into the wider open-source community. AI and tech companies have a tendency to release trained models so developers can use them without incurring the financial or environmental costs of re-training, and can benefit from their data and AI expertise. Developers should also consider reporting model efficiency and computation ‘price tags’ from training and running models (as advocated by Schwartz et. al).

There has also been progress in the techniques used to reduce the size of a model before it is deployed. A recent, promising example of this ‘model compression‘, discussed at ICLR 2020, is a variant of the ‘pruning’ technique. Neural networks (a popular AI model type often with a large footprint) undergo iterations of training and then have their weakest neuron connections are removed, effectively shrinking the models without significantly impacting performance.

Collaborating towards actionable insights

There is strong momentum behind both approaches to Green AI – using AI to solve sustainability challenges and using AI in a more sustainable way. AI has proven its commercial value but, with climate change beginning to seriously impact industries, ecosystems and livelihoods, now is the time for stakeholders across the public and private sectors to collaborate. Together we can make purposeful, lasting and necessary impact with responsible applications of digital technologies like AI.