Artificial intelligence: From massive energy consumer to problem-solving?

An image of a server Image by Dario Ruglioni onPixabay

Artificial Intelligence (AI) has the potential to drive innovation across numerous sectors. Yet, growing energy appetite related to its use raises critical questions about how much it is helping with or rather fanning climate change. A sustainability perspective requires AI to come – over even help – with a transition to renewable energy sources. Understanding this challenge is essential for ensuring that technological advancements do not come at the expense of our planet.

AI and energy consumption

One of the critical areas where AI can make a difference is its energy consumption patterns, as AI companies run energy-intensive operational activities. The energy consumption associated with AI can be categorised into two key components.

Artificial intelligence systems, especially those utilising deep learning and neural networks, necessitate significant computational resources. Model training is one of the most energy-intensive processes, as it involves iterating through large datasets multiple times, requiring vast computational power. For instance, training state-of-the-art language models can consume megawatt-hours of electricity, leading to significant emissions if powered by fossil fuels. Once trained, AI models still require energy for inference operations, which is the process of making predictions or generating outputs based on new data; as AI applications become more widespread – from autonomous vehicles to personalised recommendations – the cumulative energy used for inference can be substantial, comparable to that of small nations. In 2019, several studies highlighted the significant carbon emissions associated with training AI models. One notable study from researchers at the University of Massachusetts Amherst found that training a single large AI model can emit over 626,000 pounds of CO2, which is equivalent to the lifetime emissions of five average American cars. This research emphasised the substantial energy requirements for training advanced models, illustrating the environmental impact of AI development.

Additionally, the storage and management of large datasets contribute to energy consumption, as maintaining the infrastructure needed to store terabytes or even petabytes of data necessitate continuous electricity. Furthermore, cooling systems are essential in data centres to manage the heat generated by servers, with cooling accounting for nearly 30% of a data centre’s total energy use, underscoring the need for more efficient technologies. Collectively, these factors highlight the complex and significant energy demands associated with AI, reinforcing the importance of adopting sustainable practices in its development and deployment. The environmental impact of data centres cannot be overstated. Data centres frequently rely on fossil fuels, which significantly contribute to greenhouse gas emissions. The dependence on coal and natural gas indicates that even the most efficient artificial intelligence models can generate a considerable carbon footprint. For instance, Google and Microsoft have committed to operating their data centres on renewable energy. However, the sheer scale of their operations means that the transition will take time, and until then, the environmental impact remains a pressing concern.

Overall, these studies collectively highlight the pressing challenge of balancing AI’s potential benefits with its environmental impact, emphasizing the need for more sustainable practices in AI development and deployment. With forecasts suggesting an increase in this demand, the implications for sustainability are significant.

Balancing innovations and energy needs for sustainability

While the energy consumption of AI is a massive challenge, AI also has the potential to drive significant advancements in the area of energy sustainability, from optimising energy consumption in smart grids to enhancing climate modeling and improving agricultural efficiency. In other words: despite its energy costs, AI can help us address some of the most pressing environmental challenges. The key lies in finding the right balance. As we integrate AI into our processes, we must prioritise energy-efficient models and invest in greener technologies. Research is ongoing to develop algorithms that reduce energy consumption without sacrificing performance, paving the way for a more sustainable approach to AI.

To mitigate the environmental impact of AI, a multifaceted approach is essential to ensure that the future of AI aligns with our sustainability goals. First, developing and adopting energy-efficient algorithms can significantly reduce computational power requirements, with techniques like model pruning and quantisation helping to streamline AI models. Additionally, encouraging renewable energy adoption for data centres will diminish the carbon footprint associated with AI operations. Establishing regulatory frameworks is also crucial, as governments and organisations can collaborate to create guidelines that promote sustainable AI practices, holding companies accountable for their energy consumption. Lastly, raising public awareness about the energy demands of AI can foster a culture of responsibility and innovation, inspiring organisations to prioritise sustainability in their AI initiatives.

The transition to renewable energy and the role of AI

In the transition towards low-carbon electricity, some AI companies could be skeptical of the trade-off between the high capital intensity of renewable energy investments against the potential benefits from such investments. While the initial overall investment into renewable energy technologies, such as wind and solar photovoltaic, may be daunting, the good news is that these costs are steadily declining. In most industrialised countries, the cost of deploying and operating offshore wind, onshore wind, and solar PV has reached levels where they are now among the cheapest sources of electricity on the grid. This reduction in the levelised cost of energy (LCOE) provides a strong economic incentive to move away from carbon-intensive energy sources.

For AI companies, the falling costs of renewable energy offer a clear path forward. By investing in and integrating renewable energy sources into their operations, these companies can not only reduce their carbon footprint but also demonstrate their commitment to sustainability.

To accelerate the adoption of clean energy in AI, it is essential to foster radical technological innovations. Policies that address the emission externality, such as carbon pricing or subsidies for renewable energy, play a pivotal role in encouraging investment in research and development (R&D). These policies are key all along the entire innovation process, from upstream R&D to pilot testing and market entry, ensuring that new technologies can be scaled up and deployed as efficiently as possible. Public financial support mechanisms, such as the exemption from taxes are also crucial in fostering innovation in renewable energy projects. These mechanisms enable AI companies and other private sector actors to develop and implement cutting-edge technologies that enhance energy efficiency and drive the transition to a low-carbon economy. For instance, AI is very useful in managing electricity grids more organically and in decentralising efficiently wind and solar energy to make energy available where it is needed.

Conclusion

The AI sector holds many of the opportunities for responding effectively to climate change. By integrating sustainable methodologies into the development and implementation of AI, it is possible to leverage its capabilities to foster positive environmental outcomes. The evolution of AI does not have to conflict with our sustainability objectives. Rather, through responsible practices and a dedication to energy efficiency, we can position AI as a vital partner in the pursuit of a sustainable future. The decreasing costs of renewable energy, coupled with ongoing innovation, provide a robust economic justification for AI companies to transition away from fossil fuels and towards clean, renewable energy sources. At this pivotal moment, it is essential for all stakeholders – including governments, industries, and researchers – to collaborate in creating an AI framework that emphasises both advancement and environmental stewardship.

Kodjo Isaac Atchikiti is an alumnus of the Shaping Future Academy 2024

Kodjo Isaac Atchikiti is an alumnus of the Shaping Future Academy 2024.

Yonas Gebremichael Difer is the Director of Computational Science and Intelligent Systems Directorate and alumnus of the Shaping Futures Academy 2024

Yonas Gebremichael Difer is the Director of Computational Science and Intelligent Systems Directorate and alumnus of the Shaping Futures Academy 2024

1 comment

  1. Grace Mawusi Annor - Antworten

    Very insightful to learn how AI is linked to climate change, and the need to balance innovation and energy needs in promoting the future of sustainability.
    Great piece Isaac and Yonas. Cheers.

Leave Comment

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert