Act now to prevent the uncontrolled rise of computer science, say scientists

Cambridge scientists have laid out the principles of how computational science – which helps unlock the secrets of the environment to develop anti-cancer drugs to better understand the human race, but can have a huge carbon footprint – can make it sustainable. .

Writing in Nature Computational Scienceresearchers from the Department of Public Health and Primary Care at the University of Cambridge argue that the scientific community must act now if it is to prevent the uncontrolled rise of the carbon footprint of computational science as data science and algorithms increase in use.

Dr. Loïc Lannelongue, a research associate in biomedical data science and a postdoctoral fellow at Jesus College, Cambridge, said: “Science has changed our understanding of the world and brought enormous benefits to society. But this has brought the inevitable – and the it is not always clear – it affects the environment As scientists – as it is with people who work in any field – it is important to do what we can to reduce the amount of carbon emissions of our work to ensure that the benefits of our findings are not outweighed by the costs that damage the environment.

Recent studies have begun to examine the impact of scientific research on the environment, with an initial focus on scientific conferences and experimental laboratories. For example, the 2019 Fall Meeting of the American Geophysical Union estimates that it emits 80,000 tons of CO.₂and (tCO₂e), equivalent to the weekly air quality of the city of Edinburgh, UK. The annual carbon emissions of a life science laboratory are estimated to be about 20 tCO₂e.

But there’s one area of ​​research that’s often overlooked—and one that could have a big impact on the environment: high performance and cloud computing.

In 2020, the Information and Communication Technologies sector is estimated to have produced between 1.8% and 2.8% of global carbon emissions – more than aviation (1.9%). In addition to the environmental impact of electricity use, production and disposal of hardware, there are also concerns about water use for data centers and landfills.

Professor Michael Inouye said, “Although the environment of ‘wet’ testing laboratories is immediately apparent, the results of algorithms are not well understood and are often not considered. reducing their power, the increase in artificial intelligence and data science often means that their air can will grow exponentially in the coming years if we don’t act now.”

To solve this problem, the team has developed GREENER (Goversight, Rposition, Ecalculation, Ephysical strength and endurance, Nagreements, Education is Research), a set of principles to allow the computer science community to conduct sustainable research methods, maximizing the benefits of computer science to society and the environment.

Authority is responsibility

Everyone involved in computer science has an important role to play in making this process sustainable: individual and institutional responsibility is an important part of transparency and reducing greenhouse gas emissions.

For example, individual organizations can be instrumental in managing and developing central equipment, and ensuring that purchasing decisions take into account the design and performance of equipment procurement products. IT teams in high performance computing (HPC) environments can play a key role, both in academic pursuits and in helping scientists evaluate the progress of their work. Senior researchers can encourage their teams to think about this issue and provide access to relevant training. Funding agencies can lobby researchers to require carbon estimates to be included in funding applications.

Compare and contrast the power consumption of algorithms

Estimating and monitoring carbon emissions numerically reveals inadequacies and opportunities for improvement.

User metrics are important to understanding the environment’s impact and promoting human responsibility. The financial cost of computing is often negligible, especially in schools, and scientists can have unlimited and unnecessary computing power. Calculating the carbon footprint of individual projects helps identify the true cost of research.

Dealing with the power and consequences associated with the new partnership

Reducing carbon footprint—that is, the carbon footprint of electricity generation—is one of the most effective ways to reduce greenhouse gas emissions. This may involve transferring the calculation to countries with lower emissions, but this should be done with fairness in mind. Emissions can vary by three orders of magnitude between high- and low-income countries (from 0.10 gCO₂e/kWh in Iceland up to 770 gCO₂e/kWh in Australia).

The adoption of user devices is also a reason: one estimate found that almost three-thirds (72%) of the video streaming power on laptops comes from laptops, while 23% is used for broadcasting and only 5% for data centers.

Another important aspect is data storage. The carbon footprint of data storage depends on many factors, but the lifetime of storing one terabyte of data for a year is on the order of 10 kg CO.₂e. This issue is enhanced by the duplication of datasets so that each institution, and sometimes each research group, has a copy. Hyperscale data centers are expected to be less energy efficient, but they may also encourage unnecessary increases in computing power (‘rebound effect’).

Education and research

Education is important to inform different stakeholders. Incorporating sustainability into accounting studies is an obvious first step in reducing carbon footprints. Investing in research that will help build capacity in sustainable environmental science is a key area for donors and organizations to work on.

A recent study found that languages ​​that are widely used in research, such as R and Python, tend to be less efficient, which shows the importance of training Research Software Engineers in research teams to ensure that the algorithms used are used effectively. . There is also an opportunity to use technology more effectively by better understanding and monitoring the impact of decisions on the carbon footprint.

Dr. Lannelongue said, “Statistical scientists have a real opportunity to lead a sustainable way, but this will involve changes in our culture and the way we work. There will be a need for transparency, more awareness, better education and resources. , and better policies.

“Cooperation, open science, and access to low-carbon computing environments will also be very important. We need to make sure that solutions work for everyone, because they often have limited benefits for people, often those with low and middle incomes. countries, which suffer the most from climate change.”

Professor Inouye added: “Everyone in this community—from donors to newspapers to organizations to individuals—plays an important role and, in their own right, can make a difference. ”