By Konstantinos M. Kokkinoplitis

Artificial Intelligence (AI) is reshaping our world by changing the way we live and work, radically impacting our daily lives and transforming many areas of human activity such as medicine and entertainment.

However, despite the impressive possibilities, the issue of the ever-increasing energy requirements of AI is starting to emerge. As technology evolves more and more, the crucial question is whether we can manage the environmental and economic impacts, which threaten to widen global inequalities.

The explosive growth of artificial intelligence has created unprecedented needs for energy resources. Training and operating Large Language Models (LLMs), such as those used for language processing, image recognition and other AI algorithms, requires large-scale computing infrastructure, which consumes huge amounts of energy.

For example, a medium-sized LLM serving one million users can consume in a few hours as much energy as an average European household in an entire year. This energy intensity of AI poses a serious environmental challenge and raises questions about the sustainability of its development.

The situation becomes even more worrisome as we consider the speed at which technology is evolving. According to nonprofit organization TN Epoch, the computing power devoted to training AI models quadruples every year. It is estimated that by 2030, training algorithms may require up to 10,000 times more computing power than today. This means that training an AI model will require 200 times more energy, or about 6 Gigawatts. This equates to about 30% of the energy consumed by all data centers today. 

To meet these demands, companies are building huge data centers, which require energy not only for the model training but also for their complex cooling systems. These gigantic data centers, currently, devour resources every time we ask a simple question or ask for a translation, i.e. interact with the energy-intensive systems behind this data.

Therefore, this ever-increasing need for energy is not just a technical problem. It poses an existential problem with the use of technology in today’s world.

The first consequence of the energy requirements of AI concerns environmental costs, which can no longer be ignored. Since most of the world's electricity is still generated from fossil fuels, the increased use of which contributes significantly to greenhouse gas emissions, accelerating climate change. Despite large investments in renewable energy projects by companies like Google or Microsoft, the rapid growth of AI appears to outpace sustainability efforts. Further efforts to harness nuclear energy remain at an early stage of maturity i.e. years away from widespread application. The consequences of climate change, from increasing extreme weather phenomena to rising sea levels, make tackling the energy issue urgent.

The complex nature of the energy part of the equation leads to a second consequence: economics. Who benefits from this technology? The development and operation of AI models requires expensive infrastructure and a huge amount of energy, which is accessible mainly to countries and companies with abundant investment resources. Countries like the US and China are leading the development of technology thanks to their financial resources and strong infrastructure. Smaller countries have no such means. This contrast tends to widen the digital divide, affecting access to new technologies and opportunities.  Similarly, companies with a technological lead tend to have more resources for research into new technologies, leaving the efforts of new companies far behind in the competitive race.

Solving the problem of energy consumption by AI requires action at multiple levels. 

Researchers are working to optimize algorithms and infrastructure to improve the energy efficiency of the technology and reduce its environmental impact through research and adoption of alternative energy sources.

Governments also play an important role in encouraging the use of renewable energy and the creation of data centers in areas that need economic development and have renewable resources. An example of a successful initiative is Sweden's investment in using solar and wind power to power its data centers. These sources have significantly reduced carbon emissions. Finland also supports the creation of data centres in cold regions, where natural conditions reduce the need for energy-intensive cooling systems. These examples show how public-private collaboration can help develop sustainable solutions to the energy challenge of AI.

With the right support and concerted efforts, the digital divide can be reduced. Smaller countries can leverage AI technology, focusing on specific sectors and using renewable energy to develop it. Encouraging investments in local infrastructure and developing skills among citizens can create opportunities that allow smaller countries to participate in the global technology map. Instead of falling behind, we can foster collaboration and innovation by ensuring that AI opportunities are available to all.

AI has the potential to improve our lives, solve complex problems and increase efficiency in many areas. However, its cost, both environmental and economic, must be carefully considered. Smaller countries can leverage this technology, focusing on their local capabilities, renewable energy, and well-trained human resources. 

Greece, for example, can invest in the use of solar energy and leverage its geographical advantage to develop sustainable data centers. In addition, it can leverage its well-trained human resources and regional universities and research centres to become a pole of attraction for large companies looking to develop AI related technologies. I believe that developing know-how at local level, through educational programs and partnerships, is the way for Greece to become truly competitive in the global technological arena, reducing the digital divide and promoting sustainable development.

Building a future where AI is sustainable and inclusive requires collaboration between governments, the private sector and local communities. This partnership is key to ensuring that the benefits of AI reach everyone, not just a privileged few.

This article was first published in capital.gr on 2024/11/05.