Green AI: Building Sustainability into AI Initiatives

As AI systems grow increasingly sophisticated and powerful, a pressing concern has emerged: the environmental impact of these powerful technologies. The tech industry now faces a critical challenge – how to harness the transformative power of AI while ensuring its development and deployment align with global sustainability goals.

At the heart of this challenge lies something of an AI sustainability paradox. On one hand, the computational demands of AI contribute significantly to energy consumption and carbon emissions. Training large language models and running complex AI systems require vast amounts of processing power, often sourced from data centres with substantial carbon footprints, with Google reporting recently that its carbon emissions had soared 48% over the past five years primarily due to the rise of AI. 

On the other hand, AI offers unprecedented capabilities to optimise resource management, predict and mitigate environmental risks and drive efficiencies across various sectors, potentially leading to significant reductions in overall environmental impact.

Vincent Caldeira, Chief Technology Officer, APAC at Red Hat, describes this challenge: “AI presents a paradox in sustainability: while it significantly increases energy consumption due to its intensive computational needs, it also offers powerful tools to optimise resource management and reduce environmental impact on a large scale.”

The imperative for sustainable AI

In today’s world, the integration of sustainability into AI development is no longer just an ethical consideration – it's a business imperative: a point emphasised by Sujata Kukreja, General Counsel & Chief Compliance Officer for enterprise connectivity solutions specialist Expereo. “In 2024, sustainability and innovation go hand in hand, and the benefit of sustainable business far outweighs any initial investment or new processes required to make it work.”

Several factors are driving this integration of AI and sustainability, including growing environmental concerns, increasing regulatory pressure, evolving consumer expectations, heightened investor scrutiny and the potential for operational efficiency. To address these challenges and capitalise on the opportunities presented by sustainable AI, tech companies are adopting several strategies. 

One tactic is integrating sustainability as a core principle from the outset of AI development. Vincent Caldeira suggests a comprehensive approach: “This means integrating Sustainability Impact Assessments (SIAs) to evaluate the environmental impact of AI models during their development. This includes comparing different models and design choices based on their estimated carbon footprint and choosing the most sustainable options within the scope of acceptable model performance.”

Developing more energy-efficient algorithms and optimising AI infrastructure, meanwhile, can significantly reduce the environmental impact of AI systems. Karthik Sj, General Manager, AI at LogicMonitor, recommends specific techniques involving streamlining AI models to reduce their computational requirements without significantly impacting their performance, thus lowering energy consumption and carbon emissions. “Companies can implement techniques like pruning and quantisation to create more efficient AI models that require less computational power,” he says.

Investing in and utilising green data centres powered by renewable energy also can dramatically reduce the carbon footprint of AI operations. Many tech giants are leading the way in this area. “Google has committed to operating carbon-free by 2030,” Sujata Kukreja points out, “and uses AI to optimise the energy use of its data centres, achieving a 30% increase in efficiency.”

Conducting comprehensive life cycle analyses of AI systems helps identify areas for improvement in sustainability across all stages of development and deployment. Karthik Sj emphasises the importance of this holistic approach: “Incorporating lifecycle analysis ensures that the environmental impact is considered from development through deployment. This holistic approach allows companies to identify and address sustainability challenges at every stage of AI development.”

Fostering collaboration within the industry and supporting open-source initiatives can accelerate the development of sustainable AI practices, Vincent highlights: “Open Source Collaboration supports efficient innovation leveraging open source standards and resources such as open source AI models that are optimised for energy efficiency or open source tooling supporting efficiency throughout the ModelOps lifecycle.”

Beyond making AI itself more sustainable, companies are leveraging AI to address broader environmental challenges. Saskia van Gendt, Chief Sustainability Officer at Blue Yonder, notes: “AI-enabled supply chains can help businesses to operate responsibly and profitably via reduced waste, more efficient production, smarter transportation strategies, and reduced resource consumption.

“AI can help identify areas of operational efficiency, for predicting and preventing disruptions, reducing waste and promoting sustainable practices. AI can also improve forecasting climate events to give increased visibility for risks in the supply chain and improve resilience.”

The business case for sustainable AI

Contrary to the perception that sustainability initiatives might hinder innovation or profitability, evidence suggests that prioritising sustainable AI can offer significant business advantages. “Studies have shown that companies with strong sustainability practices outperform their peers financially in the long term,” emphasises Karthik.

The business advantages of sustainable AI are numerous and impactful, from cost savings to protecting brand reputation.

“By incorporating sustainable AI solutions into their products and services, companies can differentiate themselves in the market, attract environmentally conscious customers, and build trust with stakeholders,” Sujata elaborates. “Companies that prioritise sustainable AI practices are even more likely to draw and retain top talent in a competitive job market, as they will automatically be more attractive to increasingly conscientious and purpose-driven experts in their field.”

Challenges and considerations

While the benefits of sustainable AI are clear, companies face several challenges in implementation. These include balancing performance and sustainability, measuring impact, keeping pace with rapid technological change, aligning global efforts and managing investment costs. Vincent Caldeira addresses some of these challenges: “Implementing sustainable AI practices can lead to significant cost savings by optimising resource usage. 

“For example, energy-efficient algorithms and infrastructure reduce electricity consumption and associated infrastructure costs. Techniques such as autoscaling and in-place pod resizing in cloud environments minimise resource overprovisioning, reducing waste and lowering cloud service expenses.”

The role of industry leaders and partnerships

Tech leaders have a crucial role to play in advancing sustainable AI practices industry-wide. This involves setting industry standards, engaging in policy development, forming cross-sector partnerships, and leading by example in transparent reporting of AI's environmental impacts and sustainability efforts. 

Saskia highlights the importance of such collaborations: “Public-private partnerships can facilitate the exchange of key datasets such as weather patterns and energy usage that can be used to train AI models and develop highly advanced predictive risk models. Partnerships across technology companies, public sector and academic institutions can maximise research and development funding while harnessing the strengths of each organisation.”

And as Karthik emphasises, tech leaders should foster key partnerships with governments, academic institutions, industry consortia, and environmental organisations to advance sustainable AI practices. “Co-operating with governments ensures alignment with regulatory standards and policy development, while partnerships with universities drive research and innovation in sustainable AI technologies.”

Future outlook: The path to sustainable AI

As we look to the future, several trends are likely to shape the development of sustainable AI. We can expect increased regulation around the environmental impact of AI, driving further innovation in sustainable practices. Advancements in green computing will continue, making sustainable AI more achievable, and AI itself will play an increasingly significant role in addressing global environmental challenges. And as Vincent Caldeira explains, the industry will likely move towards standardised ways of measuring and reporting the environmental impact of AI systems.

“Collaborating with open-source communities such as the Cloud Native Computing Foundation (CNCF) can help establish common standards and reference architectures for sustainable AI, ensuring that best practices are shared and adopted widely,” he says. “The CNCF, in particular, plays a crucial role in establishing technology standards in cloud-native technology for AI, promoting efficient and scalable AI deployments. Very recently I have been engaging with the CNCF Cloud Native AI workgroup as well as the CNCF Technical Advisory Group on Environmental Sustainability on a proposal to produce a reference architecture and best practices for sustainable AI. 

“Partnering with open source, industry-driven consortia such as FINOS can drive the development of sector-specific guidelines and tools that promote adoption of energy-efficient AI tooling and also facilitate the establishment of practical standards and best practices in deploying and managing AI technology.  

“Finally, working with regulatory bodies helps to ensure alignment with environmental regulations and promotes the adoption of policies that incentivise sustainable AI practices. These partnerships collectively support the creation of a cohesive framework for sustainable AI, fostering innovation while minimising environmental impact.”

The journey towards sustainable AI is not just a technological challenge but a societal imperative. As AI continues to transform our world, ensuring its development aligns with environmental sustainability goals is crucial for the long-term viability of both the tech industry and our planet. 

As Sujata Kukreja concludes, by embracing sustainable practices in AI development, tech companies can drive innovation, reduce costs, meet regulatory requirements and ultimately contribute to solving global environmental challenges.

“If harnessed correctly and within the boundaries of regulations and ethical use, AI can be a powerful tool in supporting industries and communities to address sustainability challenges while driving innovation and positive growth.”

To read the full story in the magazine click HERE

******

Make sure you check out the latest edition of Technology Magazine and also sign up to our global conference series - Tech & AI LIVE 2024

******

Technology Magazine is a BizClik brand