We use cookies on our website to show you content we think is relevant to you, to manage the technical operations of our website, and analyse our traffic to further improve our website. We may share information about your use of our site with our social media, advertising and analytics partners. By clicking "Accept", you agree to the use of all cookies as described in our Cookie statement or "Do not accept" to only use cookies strictly necessary for the functioning of the site.
Wholesale Banking

Five ways AI can boost sustainability

21 May, 2019 |
To overview

Artificial intelligence is already changing the way we live and work. But the tech behind Siri and Alexa also has the potential to deliver big wins for society and the environment.

1. Lowering shipping emissions

Shipping is heavily polluting, accounting for 2.3% of global CO2 emissions in 2015. Moreover, with shipping growing fast, emissions could double by 2050. In a bid to do its part, ferry company Stena Line is running a pilot study using AI to reduce fuel consumption and minimise environmental impact.

The company is working with Hitachi on an AI model that will predict the most fuel-efficient way to operate a certain vessel on a specific route. The model will be a support for the captain and officers on board, and if successful it will make an important contribution to Stena Line’s sustainability target to reduce fuel consumption by 2.5% annually.

“The model simulates many different scenarios before suggesting the optimal route and performance setup. With the help of AI we are able to consider a number of variables, such as currents, weather conditions, shallow water and speed through water, in various combinations which would be impossible to do manually,” says Lars Carlsson, head of AI at Stena Line. “Planning a trip and handling a vessel in a safe and, at the same time, fuel-efficient way is craftsmanship. Practice makes perfect, but when assisted by AI a new captain or officer could learn how to optimise fuel quicker. In return, this contributes to a more sustainable journey,” adds Jan Sjöström, senior master at Stena Scandinavica.

2. Reducing road congestion

AI is certain to play a more important role on our roads in the long-term future as autonomous vehicles become widespread: it should massively reduce fatalities from car accidents. However, AI can also deliver significant congestion benefits in the short term, reducing air pollution and CO2 emissions, and improving the lives of urban residents.

Alibaba’s City Brain combines AI and real-time city data to optimise public resources such as roads. In Hangzhou, the Chinese city where Alibaba is based, the technology covers 420 square km and over 1,300 traffic lights. Since City Brain was introduced, the average travel speed on roads with automated traffic signal control has increased by 15%, reducing average travel time by three minutes, while emergency vehicle response time is 50% quicker, enabling rescue vehicles to arrive seven minutes faster. In another Chinese city, Suzhou, a pilot of bus routes increased passenger volumes by 17%. Overall, Hangzhou has moved from being the fifth most congested city in China to the 57th, according to Alibaba Cloud president Simon Hu. “Hangzhou is probably the only city that can tell you how many cars are on the street at any given time,” he adds. In September 2018, Alibaba and Hangzhou city government announced the launch of City Brain 2.0, which will optimise the city’s firefighting efforts by providing crucial information to firefighters, such as water pressure, the number and position of fire hydrants in a given area and the location of gas pipes.

3. Tackling energy efficiency

In 2016, Google and its DeepMind AI subsidiary developed an AI-powered recommendation system to improve the energy efficiency of Google’s already highly-optimised data centres. In 2018, it took this system to the next level: instead of human-implemented recommendations, its AI system is directly controlling data centre cooling. This first-of-its-kind cloud-based control system is now safely delivering energy savings in multiple Google data centres.

The system works by taking a snapshot of the data centre cooling system from thousands of sensors and feeding it into Google’s deep neural networks every five minutes. The AI then works out the best way to minimise energy consumption (while meeting safety requirements) before sending information back to the data centre where it is verified and implemented. As the AI makes a fresh assessment every five minutes, it can easily adapt to changes. “It was amazing to see the AI learn to take advantage of winter conditions and produce colder than normal water, which reduces the energy required for cooling within the data centre,” notes Dan Fuenffinger, a Google data centre operator. After just a few months of use, the system was delivering energy savings of around 30% and this will improve as more data enables the AI to make better decisions.

4. Monitoring cities and poverty

The European Space Agency generates huge volumes of satellite data – its Earth observation satellites provide about 150 terabytes every day (1 terabyte equals 1,000 gigabytes). Making sense of this information is difficult. By using AI, the data can be used to inform decisions about environmental disasters and climate change. It can also help address other UN Sustainable Development Goals (SDGs) relating to sustainable cities and communities, for example. As urbanisation accelerates in the developing world, it is more difficult to provide public health and clean water services. “As a first step towards helping people, we need to be able to map growing urban environments,” explains Patrick Helber from the German Research Centre for Artificial Intelligence. “It’s impossible to manually analyse hundreds of terabytes of data every day but AI can automatically classify land. This, in turn, allows changes in land use, such as urban growth, to be monitored automatically.”

Separately, Stanford University has used AI to tackle another UN SDG, the elimination of poverty. To track progress towards this goal, we need more frequent and reliable data on poverty distribution than is available from traditional data collection methods. The researchers’ approach combines AI with high-resolution satellite imagery to identify image features that can explain up to 75% of the variation in local-level economic outcomes. This information could potentially transform efforts to track and target poverty in developing countries.

5. Addressing gender bias

AI’s role in addressing gender equality has come under fire in recent months, following Amazon’s suspension of a recruiting tool that analysed job applicants’ resumes after it was found to be gender-biased. The problem arose because the AI models were trained to select applicants based on patterns in CVs that Amazon had received over the previous decade – most of which came from men. Other examples of gender bias in AI include voice and face recognition systems, which perform worse for women than for men.

As a recent report by EY notes, “the issue is not AI, but how humans build AI”. Nevertheless, the consequences of AI’s reliance on real-world data – which reflects social biases – are significant. Gartner predicts that by 2022, 85% of AI projects will deliver erroneous outcomes due to bias in data, algorithms or the teams responsible for managing them. This is not just a problem for gender inequality – it also undermines the usefulness of AI.

While the Amazon experiment reveals AI’s shortcomings in relation to gender equality, that doesn’t mean that researchers should give up. To make AI work properly, however, gender discrimination in society and the tech world in particular have to be tackled. Emilia Gómez is a researcher in human behaviour and machine intelligence at the European Commission’s science and knowledge service, the Joint Research Centre. She says the main problem is that “when male developers create their systems, they incorporate, often in an unconscious way, their own biases in the different stages of its creation such as data sampling, annotation, algorithm selection, evaluation metrics and the human-algorithm user interface”. The answer is fairly simple: to create “diverse teams to develop AI technologies so that these technologies are meaningful for everyone”.