Artificial intelligence (AI) has come a long way in its development since 1956 when a group of scientists and researchers first coined the term at the Dartmouth Conference. Fast forward to the early 2000s, AI has become commonplace in our homes with the introduction of gadgets like Roombas and Amazon Alexas. However, 2023 marked a major turning point, recognized as AI’s ‘breakout year,’ thanks to the accessibility of generative AI tools such as ChatGPT. Now, AI chatbots are nearly inescapable, infiltrating Google, Instagram, and even Snapchat. With the release of IOS18, Apple introduced “Apple AI,” which seamlessly integrates AI features across the entire user experience. It has become nearly impossible to navigate the internet without running into some iteration of artificial intelligence offering to help you. AI’s implications are little understood, but there are significant concerns about its societal impacts and growing energy demands.
The explosion of generative AI is truly unprecedented. Workplaces and professors alike are forced to make nuanced decisions about how to proceed with using AI in their respective fields. The implications of uncertainties such as AI errors and AI discrimination bias have begun to arise. Another little-understood yet major concern of AI is its intense energy demand.
AI training and operations depend on massive data centers, which have previously contributed only 1-2% of global energy use. However, the growing demand for AI is driving the expansion of these facilities, yielding a grim statistic regarding the carbon footprint of AI. For instance, training Chat GPT 3 was estimated to have consumed just under 1300 megawatt-hours of electricity, emitting around 522 tons of CO2– the same as the yearly emissions of around 123 gas-powered cars, before the program was even released to the public. In addition to the rising energy demand, AI boasts a significant water footprint caused by the large amount of freshwater needed for part manufacturing, electricity generation, and server cooling that make generative AI possible.
As AI grows rapidly, so does the workload on these data centers, and correspondingly so does the energy needed to fuel this technology. Amid the push towards reducing energy consumption and developing renewable energy, this increasing demand seems like it would be counteractive to the cause. However, a quick search on Google Scholar will tell you that while scholars call for more data and mindful development, they aren’t concerned about a looming energy crisis or AI-driven climate change. So what’s the catch?
It is important to consider that while AI develops, so does the efficiency of the computers that run it—so much so that AI computer chips today use 1% of the power to perform the same functions as they did in 2008, according to the International Energy Agency. As with computing technology, the cooling technologies and AIs themselves are also gaining efficiency. Even though there is still much to learn about the future energy demand of AI, this increasing efficiency objectively won’t make up for the energetic costs, but does offset the demand a bit. With limited data and the recency of AI, it’s difficult to make any concrete predictions, although there is a dialogue on whether the AI boom will stall the clean energy transition.
In fact, there is a whole field known as Green AI, which is the study of developing environmental sustainability in AI—the efforts of which are very concerned with energy consumption. While Green AI pushes for specifically developing energy-conscious AI technology, there is an element of accountability worked into all AI endeavors: the top groups that run data centers globally have previously set targets to achieve net zero emissions, and large investors must be aware of the risks associated with investments which will then appear in their portfolios. These two considerations act as somewhat of a safeguard against rampant energy negligent policies.
Concerning the sustainability risks of AI, one must recognize that the societal benefits fiercely rival the risks. Specifically in the field of climate science and sustainability, AI has been found to significantly increase the optimization of clean energy systems, have a great capacity for developing climate change adaptation strategies, and have the potential to effectively revolutionize the energy industry, among many other applications. The ability of AI to be used to solve its own problems effectively serves to minimize the environmental risk it poses.
While the environmental costs posed by the AI boom are considerable, the societal and environmental benefits AI offers—in climate science and energy fields specifically—cannot be overlooked. However, as AI continues to evolve, it is of the utmost importance that tech companies prioritize sustainability and are held accountable from the developmental stage. Only by remaining vigilant about the environmental risks of AI can we hope to strike a balance between technological progress and environmental impact.