Sustainability


performance units require substantial power to operate. And as they become more complex, their energy demands are only set to escalate, resulting in higher carbon emissions not only in the training phase but also during the operational inference phase Tough the precise energy usage figures haven’t yet been


officially disclosed, GPT-4 (the most up to date form of AI) is estimated to consume between 51,773 MWh and 62,319 MWh — the same amount of energy it takes to power 1,000 average US households over five to six years. Tis is over 40 times more than its predecessor, GPT-3. Training large AI models generates considerable carbon emissions, as researchers estimate that by 2027, the AI industry could consume as much energy annually as a country the size of the Netherlands. And the problem doesn’t end with energy usage. AI needs


hardware devices to function, and these present further environmental effects. Te production, transport, maintenance, and disposal of hardware components require additional energy use, materials, and natural resources. To exacerbate the issue, many of these materials aren’t properly recycled, creating electronic waste and further pollution.


Measuring and minimising AI’s carbon impact Despite AI’s immense impact on carbon emissions, companies can take precautions to reduce these adverse effects. Transitioning to renewable energy sources, such as solar and wind, and adopting energy-efficient practices are just the beginning of these safeguards. Te type and efficiency of the electricity sources powering data


centres also play a pivotal role in their carbon footprint. Centres utilising renewable energy sources tend to have lower emissions. Transparency in reporting energy consumption and emissions is becoming more important as stakeholders push for greater accountability and sustainable practices. Digital twins are also becoming an increasingly popular


option for reducing energy consumption. Tese tools act as replicas of physical systems and simulate real-world conditions to identify inefficiencies and reduce consumption. For instance, manufacturers might use them for modelling energy patterns to address bottlenecks, while the energy sector can optimise grid operations, predict demand, and better integrate renewable energy sources. Encouraging hybrid work is another way organisations can help


reduce their overall carbon footprint. By letting employees work flexibly from home, organisations can reduce their physical office spaces and reduce the number of emissions caused by a daily commute, leading to more sustainable business practices overall. Using smaller and more specialised models can mitigate the


environmental impacts of AI. Techniques like model quantisation can significantly reduce the size and complexity of AI models, leading to decreased computational requirements and energy consumption. Furthermore, carefully selecting AI models that are tailored to specific tasks can help avoid overengineering and unnecessary complexity. With each new generation of language models, capabilities have grown significantly, making it more practical to use smaller versions of a model to achieve similar results as previous larger models.


www.pcr-online.biz Hardware is a significant contributor to AI’s impact on the


environment. Terefore, a crucial strategy for reducing AI’s environmental influence is boosting the efficiency of the devices it operates on. And investing in energy-efficient hardware components, such as processors and cooling systems that consume less power is the first step. Dynamic power management techniques can help adjust


power consumption based on workload, reducing energy usage during idle periods. By establishing robust hardware lifecycle management plans, organisations can minimise electronic waste and extend the lifespan of their hardware, further reducing their environmental footprint. Sourcing energy from renewable sources like solar, wind, and


hydro power can help organisations significantly reduce their reliance on fossil fuels. Additionally, implementing energy storage solutions, such as batteries, can help store excess renewable energy for use during peak demand periods. Similarly, designing energy-efficient data centres with


optimised cooling systems, air flow management, and energy- efficient infrastructure can further reduce energy consumption and environmental impact. Additionally, platformisation can significantly contribute to


sustainability goals by streamlining operations and reducing the number of applications and components that need to be maintained. Platformisation refers to the integration of multiple IT capabilities such as networking, AI, and security into a unified system and is gaining traction among network providers seeking to streamline and simplify complex operations. By minimising the need for extensive physical infrastructure and optimising resource utilisation, platform-based approaches consolidate networking, AI, and security into a single integrated system. Tis reduces redundancies, lowers energy consumption, and ensures that resources are deployed where they are needed most, supporting a greener and more sustainable approach to AI and IT management.


The issue at hand AI isn’t going anywhere anytime soon. In fact, its presence is only going to increase. As AI models become more sophisticated, it’s up to researchers and organisations to find and implement sustainable solutions that can help reduce the environmental impact of AI. But it is also up to users to leverage AI in ways that create meaningful impact on society and humanity as a whole. Tis progress will outweigh the downsides of AI, especially when coupled with a continued focus on sustainability, ultimately delivering a true ROI for everyone. At Extreme, we’ve taken steps to reduce our own carbon footprint


including consolidating our labs and data centres across the United States, resulting in a 28% reduction in water usage and a 27% reduction in electricity consumption from our 2021 baseline. We’ve also introduced products such as the AP5020, a cloud-managed Wi-Fi 7 access point that optimises energy efficiency while supporting bandwidth-intensive applications like AI. By prioritising energy-efficient innovations and supporting work


environments that minimise environmental impact while improving work/life balance for employees, Extreme and the rest of the industry can help set new standards for performance and progress.


January/February 2025 | 21


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52