AI: A Problem and the Solution
My name is Alice Cloney, and I will be contributing to the RyeStrategy blog. I am a third year law student at the Hague University, studying international and European law and have completed my minor in cyber-security. My research has been focused on environmental law and its intersection with technological advancements, in addition to comparative environmental law.
The use of AI technology has expanded to all industries in recent years. With an ability to process mass amounts of data in order to assist in our understanding of large amounts of information, it is vital that industries understand the difficulties presented by AI technology. One of the biggest issues facing industries today in their investment in AI tech is the mass amount of power and space it takes to appropriately house and manage AI servers, in addition to the environmental impact this can have. However, as research into AI technology advances and solutions are generated, the issues caused by the rise of AI will most likely be solved by advancements in this same field.
Training Period for Computation
When training an AI system, especially those relying on deep learning algorithms, a large amount of electricity is used in order to ensure that the networks are capable of achieving optimal accuracy. A study from researchers at the University of Massachusetts found that these types of deep learning algorithms can emit more than 626,000 pounds of carbon dioxide. Currently MIT is undertaking research with the help of Satori, a supercomputer designed with sustainability in mind. This entails a program which provides energy and carbon feedback, enabling the user to adjust their AI model in order to save energy. Similar programs are allowing researchers to understand how to improve the efficiency of AI models, with the long term goal of lessening the amount of computation required to build an effective system.
Data Centers and Cooling
The biggest issue when it comes to implementing and monitoring AI servers is primarily the energy it takes to keep the servers running in an appropriate environment. The high-processing requirements of AI servers means it is essential to keep data center equipment running flawlessly to prevent overheating. When addressing the environmental concerns caused by the mass amounts of energy necessary to keep AI data centers running, companies are mainly looking towards AI solutions built to optimize power usage effectiveness (PUE). AI systems built to manage data centers can improve PUE through the system's ability to continuously analyze environmental changes in the center, in addition to detecting flaws, reducing processing times and resolving risk factors. The advantage of an AI system performing these tasks over traditional methods is that the system is quicker and allows for greater server optimization and performance.
In 2020 Huawei claimed that the implementation of their cloud-based service, iCooling, improved their PUE by eight percent. This service used deep learning to process sensor data, ascertain the relationship between different pieces of equipment and systems, and have the output of the cooling system match the IT load. In addition Google’s own AI system was able to help reduce the energy consumption of the data center cooling system by forty percent.
While the rise of AI has presented us with new challenges in technology and the effects it can have on the environment, these issues are quickly being addressed primarily by AI technology itself. Industries which have invested in the use of AI technology should continue to address developments in this new field, in order to limit environmental consequences. Although AI is often seen as an independent system, it is almost entirely dependent on human input -- if we emphasize sustainability, our AI systems will follow.
Learn about our affordable carbon footprint solutions for small and medium-sized businesses
Book a free strategy session to discuss your climate goals with a sustainability manager.
About RyeStrategy
Based in Seattle, RyeStrategy is a CDP-accredited, mission-oriented company specialized in carbon accounting, mitigation coaching, and climate disclosure solutions for organizations at any point in their sustainability journey. Learn how RyeStrategy helped Salesforce, Ideascale, and Wazoku achieve their sustainability goals.
From exhaustive carbon footprinting and mitigation coaching, to setting science-based targets and reporting climate data to CDP, SBTi or custom reporting platforms, RyeStrategy acts as a hands-on extension of the team, custom-tailoring services to fulfill climate disclosure requirements easily and accurately.
Meet with a sustainability specialist to learn more about RyeStrategy solutions.
