Microsoft and AI: A Sustainable Future at Stake

AI for a Sustainable Future: An Exploration by Microsoft

Technological advancements are developing at a breakneck pace, sometimes making it difficult to precisely define the future we are heading towards. In the episode titled The Shape of Things to Come, Doug Burger, head of Microsoft Research, teams up with experts from various backgrounds to examine the complex questions that artificial intelligence (AI) poses to technologists, policymakers, business leaders, and other stakeholders. The primary goal is to enhance the common understanding necessary to build a future where the transition to AI results in a net benefit.

In this episode, Doug Burger is joined by Amy Luers, head of sustainability science and innovation at Microsoft, and Ishai Menache, optimization researcher at Microsoft Research. Together, they explore how AI can both contribute to climate change and help address it, while emphasizing the importance of distinguishing hype from real data and understanding the true impact of AI. Although data centers account for only a small share of global emissions, their rapid growth raises concerns about local infrastructures. However, AI offers powerful tools to optimize complex systems and accelerate climate solutions. The discussion presents AI as a crucial yet double-edged technology that must be carefully guided to support a sustainable future.

Podcast The Shape of Things to Come

Doug Burger opens the episode by restating the ambition of The Shape of Things to Come, the Microsoft Research podcast he hosts: to explore the capabilities of AI, delve into the fundamentals, and reflect on how these capabilities will transform the world, for better and for worse.

To discuss the future of AI and sustainability, he welcomes two experts. Amy Luers, Senior Global Director of Sustainability Science and Innovation at Microsoft, works at the intersection of sustainability, technology, and science. Ishai Menache, a research lead at Microsoft Research, is a globally recognized expert in optimization.

The conversation starts from the way technology can optimize systems. Can AI help combat climate change and serve sustainability, and what challenges does it bring in return? For Doug Burger, the point is to get to the heart of the matter, because that is what will determine the shape of things to come.

The host makes no secret of how close the subject is to him. He says he has been a climate advocate since the 1990s and that the climate crisis worries him greatly, while noting that Amy Luers has dedicated her career to it.

He then sets out the context. The transition to AI is under way and the tech industry is building massive computing systems, mega data centers. Their possible impact on the climate is raising many concerns around the world, and local communities will be discussed as well. Doug Burger therefore wants to examine the facts: what does this build-out actually mean, and what will its real impact be? "Let’s separate data from hype," he suggests, before announcing that the discussion will also cover the opportunities ahead, which is why Ishai Menache is there.

Amy Luers begins by describing her role. She leads sustainability science and innovation within Microsoft’s sustainability team, which means working with teams across the company, worldwide, and with Microsoft Research, to shape and inform sustainability solutions for Microsoft but also for the rest of the world. Part of that job is leading the group’s strategy on AI and sustainability.

"I have worked on sustainability and climate my entire life," she sums up. Her career runs through the tech sector, at Google, and through the White House, where she worked at the intersection of the CTO’s office and environment, resources, and energy issues. She also led an international research network linked to the UN and focused on sustainability.

It was at Google that she started to think about the power of computing and digital tools for transformation. After leaving, she was led to work at this intersection. When she took the helm of the sustainability research network Future Earth, she says she really felt the need to think about innovation and digital technologies in this field.

About ten years ago, she recalls, the sustainability science network was "a bit reluctant" to consider AI and technologies. She then launched a global initiative, Sustainability in the Digital Age, to bring together the digital technology and AI community, in Montreal, where that community is large, and sustainability scientists from around the world. She began to think about the potentials and the risks, and led a large international study to establish a research and innovation agenda in this field.

That experience changed her approach. She moved from the idea that "big computing can help," which she was focused on at Google, to the role of AI and machine learning in this area.

Doug Burger then introduces Ishai Menache as a globally recognized expert in machine learning and optimization, who has published extensively, including in Harvard Business Review, and who is well known in his research community. He also points to his significant impact on Microsoft, and describes him with humor as "a bit of a polymath and sometimes a bit intimidating."

Ishai Menache explains that his background is first of all in engineering. His graduate studies were in machine learning and reinforcement learning, then in distributed optimization and game theory, on the more theoretical side.

His story with the cloud begins around 2009, during his postdoc at MIT [Massachusetts Institute of Technology], when the cloud was on the rise. He was first interested in its economics, and in particular in how to price it. He discovered Microsoft Research at that time, when a new type of lab, MSR New England, was opening near MIT. Beyond the economic aspects, a more fundamental question fascinated him: how to use resources more efficiently.

That is what brought him to Microsoft Research in 2011. First a consultant at MSR New England, he moved to Redmond that same year to join the Extreme Computing Group, a lab that was dealing with the future of the cloud. Doug Burger was part of it too, so the two men have known each other for a while.

The researcher describes the angle he chose at the time. On one side, many systems specialists were thinking about cloud infrastructure. At the other extreme, theorists were thinking about the next wave or trying to innovate in the algorithm space. "I think what was missing was a bit of a bridge between algorithms and cloud infrastructure," he explains. He found a very interesting niche there, for himself and later for the group he founded in 2019.

Ishai Menache recently announced a system called OptiMind. Doug Burger, who was excited enough about the announcement to post about it on LinkedIn, asks him what the system does and why it has attracted a lot of attention, with the idea of then delving into optimization for the audience before coming back to AI.

The researcher starts by taking a step back and defining mathematical optimization: a way of using mathematics to make the best decisions when there are many choices and certain limitations.

Every optimization problem, he explains, starts with a description of the problem to be solved. Then comes a set of decisions, which are the variables in mathematical terms, and then an objective, meaning what one is trying to optimize. It may be a matter of maximizing something, like revenue, or of minimizing costs: there are different types of objectives. Then there are the constraints, which rule out doing whatever one wants. These are capacity limits in the case of the cloud, for example, or other factors to consider in order to arrive at the best possible decisions.

Doug Burger then offers a simple and deliberately light example: a complex commute to work, a usual route that is blocked one day, and brakes so worn that the driver can only stop twice. The framework described by Ishai Menache might be able to determine which route gets the driver there while saving the most fuel. "Exactly," the researcher confirms, adding one more possible constraint: the driver may also not want to pay tolls, for some reason, which narrows the possible routes accordingly.