The hype around AI is jumping from one industry to another as celebrities, scientists, and other professionals are using it to speed up their usual routines.
While some environmentalists believe that AI can fight climate change, it's important to understand first that it also has a problem: its growing carbon footprint.
Should We Worry About AI's Carbon Footprint?
When people started to invest in powerful computers for crypto mining, some experts believed that it should stop as early as possible. Investors and traders indeed gain an advantage through this, but they are not aware of its impact on the environment.
The same goes with artificial intelligence which is also considered "environmentally taxing." It might appear to be an instant remedy for all the tedious tasks done in the lab, school, or institution. In reality, continuous use is just the tip of the iceberg.
According to The Guardian, GPU-powered AI tools are capable of handling huge amounts of calculations in a matter of seconds. The GPUs are also used in crypto mining so that sparks a debate about AI's environmental aftermath.
As for Hugging Face's researcher in ethical and sustainable AI, Sasha Luccioni, the study should dive deeper into AI's carbon footprint.
"Fundamentally speaking, if you do want to save the planet with AI, you have to also consider the environmental footprint [of AI first]. It doesn't make sense to burn a forest and then use AI to track deforestation," Luccioni says.
Quantifying AI's Carbon Footprint is Difficult
While the environmental footprint that AI causes is possible to measure, Luccioni believes that it will not be a walk in the park. She adds that there are several reasons why it will be difficult to quantify such aspects.
For instance, chipmakers won't always be transparent in sharing the details about the energy use of their products.
In some way, people also think that AI does not count as material, that's why they take it out of the equation when it comes to measuring its carbon footprint.
Luccioni says that people can imagine AI as ephemeral which is far from what computers have which is a physical form.
The researcher continues that this kind of ephemerality has been existing for the end users. While people think that carbon-emitting vehicles are harmful to the environment, they do not give the same treatment as AI since it's just a cloud-based platform.
To solidify future research about AI's environmental footprint, an academic study entitled Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models was released.
According to the research, the training of GPT-3 was able to use 1,287 MWh. The authors said that this can be compared to flying between San Francisco and New York 550 times.
By 2025, the energy consumption of AI products will surpass the entire human workforce. For the next five years, 3.5% of the total electricity consumption in the world will be accounted for by machine learning and data storage.
