Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.
COP28 UAE: AI and IT at Odds in Sustainability
At COP28, the ITU said ITCs can meet all 17 of the UN’s sustainability goals. But can IT really do that in the face of AI’s insatiable demand for more energy?
7 Min Read
Participants walk on Al Wasl Avenue on the last day of COP28, UN Climate Change Conference.Michal Busko via Alamy Stock
At a Glance
- The three highest-emitting sectors -- energy, materials, and mobility -- can reduce emissions by 4%-10% by 2030.
- One of the best ways to fine-tune current sustainability efforts and find new ones is to collect and analyze relevant data.
- AI is a known energy hog. Although precise energy usage varies, no one argues that the numbers are not sustainable.
COP28, the United Nations Climate Change conference recently held in the UAE, became contentious over universally declaring the fossil fuel era dead. While China and Russia worked to hobble efforts on reducing dirty energy use in general, oil-producing countries like Saudi Arabia vigorously objected to a multi-country deal to phase out fossil fuels in particular. The conference wrapped with a watered-down document of voluntary promises hailed as “the beginning of the end” of the fossil fuel era.
Much ado ensued as to the historic solidarity of 200 parties signing a very weak agreement. Meanwhile, the International Telecommunication Union (ITU), confidently declared information and communication technologies (ICTs) can meet all 17 of the UN’s sustainability goals. But can they, given the growing demand for the biggest energy hog in the room: AI?
High Hopes for IT to Save the Day
The ITU, which is the United Nations specialized agency for ICTs, is enthused about ICT contributions to the 17 United Nations Sustainable Development Goals (SDGs). COP28 watchers are generally more confident in the outcomes of ICT efforts than those likely to result from the newly inked fossil fuel agreement.
All 17 SDGs are presented alongside efforts ICTs can undertake to meet them on the ITU website. Several of them also align with suggestions on how to reduce an organization’s carbon footprint and energy costs. The fact that these efforts have multiple winners helps keep the momentum going. The same cannot be said for taking a hard approach to phasing out fossil fuels when so many countries make their fortunes from oil.
While there are numerous ways for IT and ICT professionals to don their capes and join the fight to save the planet, there’s only room for three in this article.
Lessen Data Center Impact
The World Economic Forum reports that the three highest-emitting sectors -- energy, materials, and mobility -- can reduce emissions by 4%-10% by 2030 by accelerating the adoption of digital technologies and by 20% by 2050 if they’re used at scale. Other industries can presumably see similar emission reductions and energy cost cuts by following suit. The primary question is how?
John Clinger, director, energy and sustainability at ICF, a global consultancy and technology services provider, offers the following how-to suggestions. His role is to “support the US Environmental Protection Agency by overseeing a team of engineers, analysts, and subcontractors developing ENERGY STAR technical eligibility program requirements for over 75 product categories.” Further, he helped develop the EnergyStar resources linked below.
Clinger suggests IT and ICT professionals take the following steps:
Temperature setpoints: Modern IT hardware is designed to operate reliably up toward the edge of ASHRAE recommended thermal guidelines, roughly 81°F. IT operators should strive to operate their data center spaces closer to this target temperature, but not to venture past it or fan power in the IT hardware itself can begin to outweigh energy savings on the facilities side of the equation. Every degree you can raise your data center results toward that 81°F target results in a 4%-5% energy savings per 1°F raised.
IT hardware power management: Your IT equipment comes with various levels of power management features that are often enabled when shipped. Have your team take a look at what level of power management is appropriate for your specific workloads and use those features as appropriate to greatly reduce IT hardware power when products are idle. Avoid turning them off blindly as these features can save you 20% or more on your IT hardware energy use when implemented properly.
Hardware utilization: Commodity enterprise IT hardware is typically designed to operate reliably up to utilization levels of 50% on average, but many enterprise users are still operating in the 20-30% average utilization range across their fleets. Using DCIM software and other tools to identify underutilized hardware and virtualization and other similar techniques to better leverage that underworked hardware can lead to less energy use across the hardware fleet. It can also result in less IT hardware needed for a given load, potentially resulting in further reduced hardware, cooling, and software licensing costs.
Embedded data centers: Whenever possible, consolidate and eliminate smaller server rooms and closets from your operation, moving that compute capacity to an in-house purpose-built data center environment and/or cloud resources. This is often the single largest bang for your buck potential for improvement on the facilities side of IT as operating IT hardware in spaces not designed for that load is typically very inefficient.
Finding and Using Data to Improve Sustainability
One of the best ways to fine-tune current sustainability efforts and find new ones is to collect and analyze relevant data. But again, it can be difficult to determine where to start.
“Start with data like temperature trends, humidity levels, and pollution statistics, and then correlate with internal IoT data from environmental sensors and energy meters. This integration can reveal how external environmental conditions impact internal energy usage and efficiency,” advised Doug Roberson, COO of Shelly USA.
Most organizations are familiar with analyzing historical energy usage and equipment emissions to predict future usage and where to offset usage to coincide with downturns in demand. But combining internal and external data can offer ways to improve on these and more.
“Insights from climate and IoT data can be used to develop and refine sustainable IT strategies, including optimizing data center operations, server cooling systems, and overall IT infrastructure layout to minimize energy use and carbon footprint,” Roberson said.
But it’s not just about analyzing data. Organizations need an actionable plan, too. When it comes to implementation, policy is everything.
“Empower IT departments to make data-driven decisions about policy making. Policies could include energy-saving settings on all IT equipment, optimizing work-from-home versus in-office work, and green procurement practices,” Roberson said.
Require Vendors to Be Green
“Even as tech’s carbon footprint grows, 59% of sustainability managers say vendors’ sustainability is likely to be overlooked during the vendor selection process,” said Shawn Rosemarin, vice president of research and development for customer engineering at Pure Storage.
The primary reason purchasing agents aren’t requiring vendors to be actively practicing sustainability, according to Rosemarin, is that “electricity consumption and environmental sustainability is not their responsibility or a pressing issue in their mandate.”
This is a problem that can be fixed in a way that produces profound benefits.
“Integrating these considerations from the outset can ensure that technology procurement aligns with broader sustainability goals. Recognizing IT as the epicenter of sustainability, organizations should engage sustainability managers before the technology purchasing process begins, ensuring their input shapes decisions to prioritize efficiency, scalability, and simplicity in IT infrastructure,” Rosemarin said.
But there is also one last piece of advice in fully evaluating vendors for environmental impact.
“Sustainability needs to look beyond the initial pilot phase, factoring impacts to a full-scale implementation,” Rosemarin said.
AI as Sustainability Saint and Saboteur
AI is a known energy hog. Although precise energy usage varies between models, no one argues that the numbers are not sustainable.
“In 2019, researchers found that creating a generative AI model called BERT with 110 million parameters consumed the energy of a round-trip transcontinental flight for one person,” wrote Kate Saenko, associate professor of Computer Science at Boston University, in The Conversation, citing "Energy and Policy Considerations for Deep Learning in NLP."
“Researchers estimated that creating the much larger GPT-3, which has 175 billion parameters, consumed 1,287 megawatt hours of electricity and generated 552 tons of carbon dioxide equivalent, the equivalent of 123 gasoline-powered passenger vehicles driven for one year. And that’s just for getting the model ready to launch, before any consumers start using it,” Saenko continued, citing "Carbon Emissions and Large Neural Network Training."
ChatGPT now runs largely on GPT-4, which is an even larger model. It’s easy to see how fast that carbon footprint grows when you consider that ChatGPT captures 60% of market share according to a study by Sujan Sarkar at writerbuddy.ai. Compounded with many different types of AI on the market and all seeing a strong and sustained demand, the overall impact rises staggeringly high, fast.
But AI can also help IT in reducing environmental impact. And it can do so in many different ways.
“Consider using AI-powered digital twins to bring together operational, weather, and quality data to optimize key processes like manufacturing production,” said Florian Zittmayr, team lead for data science at Wienerberger AG, the world’s largest producer of bricks, which is using digital twin technology (a form of generative AI) to lower its use of natural gas, reduce greenhouse emissions, and improve product quality.
According to Zittmayr, Wienerberger “collects data from a wide range of sources, from IoT edge devices and sensors throughout its factories to environmental data on weather and humidity to product spot checks during and after production. All this data reveals fluctuations that can lead to inefficiencies and unnecessary energy use.”
IT and ICT professionals will need to be proactive on curbing AI’s environmental impacts while also accelerating its insights on sustainability actions.
About the Author
You May Also Like
More Insights
