Greek MIT Professor Seeks to Tackle Data Centre Energy Consumption

Greek computer scientist Christina Delimitrou, an associate professor at the Massachusetts Institute of Technology (MIT), is using machine learning to improve the efficiency, security and reliability of large computing systems, aiming to reduce energy consumption and the need for new data centre infrastructure as artificial intelligence drives up global electricity demand.

Based in Cambridge, Massachusetts, Delimitrou recently secured tenure as an associate professor at MIT. According to a profile published by MIT News, her research team is developing ways to make better use of existing computing resources, addressing a growing challenge for the global economy and the environment as data centres consume increasing amounts of electricity.

Speaking to the Athens-Macedonian News Agency (AMNA), in an interview with journalist Alekos Lidorikis, Delimitrou described the sustainable development of artificial intelligence as one of the most significant technological challenges of the coming years.

“Artificial intelligence is changing the way we work and interact with technology at an unprecedented pace. This progress, however, comes at a significant cost that remains invisible to the end user: the enormous computing infrastructure and energy required to make it work,” she said.

Data centres have become critical infrastructure for modern society, but their growing energy requirements create significant environmental challenges.

“One of the most important questions we must answer in the coming years is not only how to make artificial intelligence more powerful, but also how to ensure that its development remains sustainable,” Delimitrou said.

Her team at MIT is addressing this challenge by rethinking how large computing systems operate. Research has shown that a substantial proportion of available computing capacity often goes unused, even as these systems continue to consume significant amounts of electricity.

“We use artificial intelligence itself to improve the efficiency of this infrastructure, allowing us to support more applications with fewer resources without sacrificing performance or reliability,” she said.

“The real innovation lies not only in creating increasingly powerful technologies, but also in our ability to make them more efficient, accessible and sustainable.”

Tackling unused computing capacity

During her postgraduate and doctoral studies at Stanford University, Delimitrou worked with Professor Christos Kozyrakis to investigate the performance of large-scale computing systems. They found that many systems operated at only 15% of their capacity despite growing demand for computing power.

Delimitrou attributes much of this inefficiency to software. Better management of existing resources could reduce electricity consumption and limit the need to build additional facilities.

Her research team develops machine-learning methods to automate resource management in cloud computing environments. The team also redesigns computing systems and server architectures to improve their efficiency.

One example is Seer, a tool that uses deep-learning techniques to predict and prevent problems in online applications before they occur. The technology helps reduce delays and service interruptions in applications such as video conferencing and online music streaming.

By anticipating potential problems, Seer aims to improve reliability while helping computing systems make better use of their available resources.

From the National Technical University of Athens to MIT

Delimitrou grew up in a city in northern Greece in a family with a scientific background. Her mother worked as a chemical engineer, while her father was a pharmacist. Both encouraged her interest in mathematics and the natural sciences.

She studied computer engineering at the National Technical University of Athens (NTUA), where she began exploring computing resource management. She continued her studies at Stanford University and pursued her academic career at Cornell University before joining MIT’s faculty in 2022.

She now works in MIT’s Department of Electrical Engineering and Computer Science (EECS) and belongs to the Computer Science and Artificial Intelligence Laboratory (CSAIL).

Reflecting on her academic journey, Delimitrou told AMNA that her education in Greece played an important role in preparing her for an international research career.

“Having started my studies at the National Technical University of Athens, I feel particularly grateful to the professors and fellow students who played a decisive role in my scientific development,” she said.

She added that the emphasis on critical thinking and the people with whom she had the opportunity to collaborate had equipped her to pursue research abroad.

Delimitrou believes Greece has exceptional scientific talent among both academics and students, and that the quality of research conducted at Greek universities deserves greater international recognition.

“I hope young scientists in Greece will continue to set ambitious goals, challenge what we take for granted and believe they can make a meaningful contribution to the greatest scientific and technological challenges of our time,” she said.

Adapting computing systems to new demands

The rapid evolution of digital applications is changing the way engineers must design computing systems. Many modern services consist of smaller components that operate across different servers, creating new demands for both software and hardware.

Delimitrou’s team combines machine learning with computer architecture design to develop solutions that help systems handle these increasingly complex workloads more efficiently.

Alongside her research, she places considerable importance on teaching. One of her favourite courses at MIT is Computation Structures, which attracts approximately 350 undergraduate students each semester.

In comments published by MIT News, she explained that she aims to foster independent thinking and creativity by encouraging students to find solutions for themselves.

Her work highlights a potential way to address the growing energy demands of artificial intelligence: improving the efficiency of existing computing infrastructure rather than relying solely on the construction of new data centres.

As demand for computing power continues to grow, making better use of existing resources could help reduce electricity consumption, ease pressure on power grids and support a more sustainable digital economy.

Stay connected to Greek City Times for Free on Google News
Click NOW
Stay updated with the latest news from Greece and around the world on greekcitytimes.com.
Contact our newsroom to share your updates, stories, photos, or videos. Follow GCT on WhatsApp, Google News and Apple News.