On June 5 for World Environment Day, GSF hosted a live virtual event for tech enthusiasts interested to know how software can help save the planet. Attendees joined GSF experts to hear about the current state of green software, the challenges and opportunities ahead, and the best practices and tools to make code more eco-friendly. Over 200 practitioners (a number that would have once shocked us, but not any more) across the globe participated and raised questions about a broad set of issues, from cloud computing to tools for sustainable AI.
Our expert panel included some of the most influential voices in the green software movement, such as:
Asim Hussain, Chairperson & Executive Director at GSF and Director of Green Software & Ecosystems at Intel
Anne Currie, Green Tech Advocate at Container Solutions and Community Chair at GSF
Tamara Kneese, UX Research Leader & Strategist and Lead Researcher at GSF
Pindy Bhullar, ESG Chief Technology Officer at UBS and Ph.D. Researcher
Our panelists kicked-off the event by introducing the Green Software Foundation and the State of Green Software report, before diving into the insights they found most compelling.
Watch our Green Software Revolution event on YouTube.
What GSF Stands For
The Green Software Foundation focuses on how we can build software with zero harmful environmental effects. The GSF is the global community of developers and organizations committed to making software more sustainable. Together we are building a trusted ecosystem of people, standards, tooling, and best practices for Green Software.
The State of Green Software Report
The State of Green Software report is a digital publication and centralized hub offering primary and secondary data, knowledge, and perspectives on green software, empowering developers, designers and solution architects at any level to secure funding, launch, and scale green software projects.
The report looks at the levels of awareness and adoption of green software, the barriers and enablers of adoption, and a deep understanding of what developers are working on in the space. It also attempts to comprehend the intersection between academic knowledge and tooling development. It has 30 sections with easy-to-digest insights. The panel handpicked a few of those for discussion in the web event.
A full recording of the event can be found here.
Expert's Top-Picked Insights
Carbon aware software is central to decarbonization
Carbon awareness describes the ability of software to location- or time-shift workloads. The better we can line up workload processing with the availability of clean energy, the more carbon aware our application is and the fewer carbon emissions it causes. One of the most popular open source projects within the GSF is the Carbon Aware SDK, which enables the integration of carbon awareness into software.
“You don't put your clothes out to dry when it's raining; you don't get upset, you just put them out when it's sunny. You simply respond to nature. As technologists, we are often very detached from the physical world. The interesting aspect of carbon aware computing is making your software do more when the sun is shining, and the wind is blowing; you’re adapting kind of to the natural order of the world,” so Asim Hussain.
Anne Currie adds: “When it comes to tooling, I will say that you must be able to sell it within your business. If it's fundamentally misaligned with your business objectives, you will have a hard time making a change, no matter how good the tools are. So you need to find a way to align green software with a high-priority business goal, which is often: Going live as fast as possible. You are looking for tools that enable you to be green and go fast simultaneously.”
Responsible AI is green AI
Tamara Kneese shared that when looking at the environmental impact of generative AI, we will need a long-term lens, not just looking at the carbon cost of training the model but the entire life cycle. She added: “Intertwining responsible AI with green AI is of utmost importance, something several people in our survey pointed out in their comments. As generative AI grows, we will have to pay even more attention to this aspect.”
Pindy Bhullar followed and shared: “There are considerable impacts on several layers, wider topics not just the environment, social impacts such as labor, and governance considerations. Responsible AI fits around the ESG topic as a whole. While the energy perspective is massive, we must watch the impact on society. It will be challenging but interesting to see how we can solve it.”
Will AI be used inefficiently and raise energy consumption?
Hardware efficiency and speed are so high that vast amounts of data can be processed. In that regard, our code has become less efficient. AI is the culmination of that. It uses vast amounts of energy due to the high volume of data it processes. If AI is used responsibly for high-value practices, there are few concerns. But unfortunately, we can’t be sure about that. The fear is the excessive use of AI for low-value computing.
63% of CEOs do not rate sustainability as a top priority
“The interesting part is that there is investor, market, and regulatory pressure on C-suite and leadership to advocate for sustainability. Yet I would argue that a lot of the internal pressure comes from employee-led groups and passionate volunteers. If you look at where resources are being deployed, that will tell you about real commitment by ICT leadership,” said Tamara Kneese.
Asim Hussain added: “We have seen with multiple organizations the real transition, almost overnight, comes when executive pay is somehow linked to sustainability.”
Decarbonization alone cannot make software green
A recent study by the University of California Riverside and the University of Texas Arlington shows that training an AI model in a data center consumes around 700,000 liters of fresh water. “This is a new angle on the environmental implications of software, broadening the scope and not only looking at energy and carbon emissions but also at water consumption. The consequences can be severe, especially in areas predisposed to water shortages and droughts,” said Pindy Bhullar.
AI training workloads are generally not latency sensitive, so there is significant flexibility as to where to run them. The challenge is aligning carbon efficiency with water consumption. Asim added, “You may be trying to time-shift or location-shift a workload, only to land in a data center in a dry place. Now you have transformed the carbon problem into a water problem.”
If you want to dive deeper into green software, here are a few interesting resources that we shared during the panel discussion:
State of Green Software Report
Green Software Practitioner Course
This article is licenced under Creative Commons (CC BY 4.0)