By K V Ch Madhu Sudhana Rao
The writer is an AI researcher based in the USA
Advances in Artificial Intelligence (AI) are generating excitement worldwide. Better healthcare, more effective research and education, and improved productivity are just some of the anticipated benefits. Around the world, billions of dollars are being invested in AI by both governments and the private sector. Global competition to build the fastest computers and the most sophisticated AI systems is on the rise. While these developments are undoubtedly promising, they come with an uncomfortable trade-off: the energy costs of powering these machines.
The environmental impact of the rapid proliferation of data centres is alarming. This includes land clearing, water and energy depletion, rising demands on the power grid, and other environmental consequences. Increased investments in tech-related infrastructures are often justified by policy-makers as critical to the digital economy. However, when the consequences that tech-related developments have on the environment are ignored, and communities are excluded from decision-making, advances in tech quickly become a justification for wanton environmental destruction.
The physical infrastructure required to power AI systems has an enormous environmental cost. The same applies to data centers. Harvesting and processing the materials required to power data centers disrupts ecosystems and contributes to land degradation and erosion. Deforestation also removes crucial carbon sinks and wildlife habitats. The contradictions are obvious. AI is being developed and deployed to help solve the climate crisis, track and monitor biodiversity loss, and improve the efficiency of using and managing the world’s natural resources. Yet the supply and infrastructure needed to provide these services put significant stress on ecosystems around the world.
Not every data center is bad. The environmental impacts of data centres depend on location and several factors, including the center’s design, how it’s cooled, and the energy sources used. The problem is that the expansion of the world’s computing infrastructure may outpace the systems in place to assess and manage its environmental impact.
Data centres can expand for valid economic reasons. Well-designed and managed data centres create new, high-tech employment, improve a country’s economy, and strengthen its digital and networking capacities. From a national and business perspective, data centers offer an opportunity to diversify an economy, improve service offerings, and attract new investment.
What happens in one country should not matter more than what happens in other countries. There should be clear rules and regulations that foster innovation, but also ensure that the environment isn’t being harmed and costs are not shifted to the public. This is especially pertinent to countries with aggressive digital agendas. Nations trying to attract data centre investment must evaluate the long-term cost-benefit of energy and water use, as well as land availability.
Several ways can minimise the environmental impact of AI infrastructure. More effective use of AI hardware and better environmental monitoring are also within the realm of current technological possibility. Companies in the AI sector should provide detailed, relevant, and accurate information on their energy and water consumption. This information should be made available to the public to help governments and local communities evaluate the true impacts of AI infrastructure.
Artificial Intelligence may resolve global challenges. However, the field can only provide so much. A society’s goals must determine the direction of human affairs and science. Aiming too high in our quest for digital intelligence may deplete essential resources, disrupt natural ecosystems, or neglect the needs and/or desires of local communities. Eventually, societies must consider whether the costs for such goals are worth the outcome.
Although we must develop sophisticated, cleaner, more effective technologies, our goal should be to preserve and improve the natural world. Thus, if there are trade-offs, it is better to sacrifice technology than to sacrifice ecosystems. The future of AI cannot be at the cost of the natural world.
