Training AI Models Might Be Less Environmentally-Friendly Than We Think

There is a lot of focus and emphasis placed on tech that is sustainable. This is because our resources are finite, meaning that there will eventually come a point in time where a certain resource is no longer available. Plus, we also need to think in the long-run, where our actions today will no doubt have an impact on the generations to come.

Now, AI is without doubt part of our future where we have seen how it can be used in a variety of industries, such as health, where it could be used to detect things a human doctor might miss. However, as beneficial as AI could be, it seems that training AI models might be quite bad for our environment.

This is according to a recent paper published by researchers at the University of Massachusetts, Amherst where they have found that the process of training an AI model can emit more than 626,000 pounds of carbon dioxide equivalent. This is estimated to be about five times the lifetime emissions produced by an average American car, which also includes the process of manufacturing the car itself.

According to Carlos Gómez-Rodríguez, a computer scientist at the University of A Coruña in Spain who was not involved in the study, “In general, much of the latest research in AI neglects efficiency, as very large neural networks have been found to be useful for a variety of tasks, and companies and institutions that have abundant access to computational resources can leverage this to obtain a competitive advantage. This kind of analysis needed to be done to raise awareness about the resources being spent […] and will spark a debate.”

While this does not mean that the research of AI should be halted, but rather, the paper hopes to raise awareness of the environmental impact that training AI models might have, and maybe inspire other researchers to develop more efficient hardware and algorithms for the future.

Tyler Lee
Categories: Computers, General, Green
Tags: AI (Artificial Intelligence)