Despite growing concerns over the high energy consumption of AI infrastructure, the Center for Data Innovation (CDI) maintains that it does not pose a significant threat to the Earth’s climate. This belief comes in the face of the high energy utilization of GPUs needed for training large language models. The CDI, instead, suggests that AI can leave a positive impact on the climate by replacing inefficient processes and optimizing others. This, in turn, can lead to a reduction in total energy consumption.
In support of this argument, a study by Cornell University has indicated that the carbon emissions produced by AI when writing a text page are significantly lower than those produced by a human using a standard laptop. This calculation excludes the CO2 produced during the initial training of AI.
To further bolster the case for AI, the CDI report states that AI could achieve energy efficiencies similar to a smart thermostat. This can be achieved by predicting grid demand, optimizing the use of water and fertilizer in farming, or monitoring methane emissions from satellite data.
The CDI does not dismiss the environmental impact of AI, but instead urges the development of standards for measuring the power consumption and carbon emissions of AI workloads. The organization promotes voluntary reporting and recommends government investment in AI for decarbonization.
Assessing the positive and negative impacts of certain factors can be challenging, but often the real issue lies elsewhere. For instance, running climate models consumes significant energy, which is considered wasteful if the ‘science is settled.’ Instead, this energy could be used to improve useful weather models like the ECMWFs.
AI has the potential to improve weather and climate models, make agriculture more efficient, among other beneficial outcomes. However, the misuse of AI, such as in creating deepfakes, is seen as a waste of energy. Therefore, it is crucial to strike a balance in its use to ensure that AI’s benefits outweigh its potential drawbacks.
Science4Data is committed to cut through greenwashing and measure real impact. Join the journey to a sustainable future. Your actions matter.
