What Impact Does Generative AI Have On The Environment?
Artificial Intelligence has created quite a buzz over the past several years. We’ve talked about its prospects and immense benefits, and even its potential consequences regarding ethical use, job displacement, the elimination of certain trades, and more. But at its current stage of development, what aspect do many of us not consider? What do we know about AI’s environmental impact?
Before you ask your go-to AI tool to generate a professional headshot for your LinkedIn, let’s take a moment to review some current research. Is it worth the footprint? Furthermore, what can we do to use this exciting, fresh tech as mindfully as possible?
Here’s everything you need to know about generative AI and its implications on the environment. This article is not intended to deter the use of AI; rather, it is an incentive to start an open conversation about how we can both reap its benefits and ensure that safeguards are in place for the sake of our planet.
The Environmental Impact Of Generative AI: Figures And Statistics
It’s safe to say that when the first AI model became available for public use, many people were excited, while others were not. Witnessing technological breakthroughs in one’s lifetime is an unparalleled experience, and this has led many people to give chatbots and GenAI tools a try. But what preceded their launch, and what does it mean for the environment? Let’s talk numbers.
Any AI tech or tool starts with training. In simple terms, this means feeding the AI model with the data required to provide answers. The training of a single AI model consumes thousands of megawatt hours—GPT-3’s training reportedly required 1,287 MWh, emitting 552 tons of CO₂. [1] That’s equivalent to 123 gasoline-powered passenger vehicles driven for a whole year. [2] In addition, as the electricity demands fluctuate during training phases, they require power grid stability solutions that often use diesel-based generators. Combining that with the fact that energy usage continues beyond training, through deployment, user interaction, and iterations, it’s safe to say that the environmental footprint is massive.
Moreover, data centers are central to training and running AI models. They typically contain tens of thousands of servers each and are rapidly multiplying to keep up with the increasing AI demand. From 2022 to 2023, North American data center energy needs doubled from 2,688 MW to 5,341 MW. In 2022, global use reached 460 TWh. By 2026, global data center consumption could amount to 1,050 TWh, marking them the 5th among the world’s largest electricity consumers. [1]
What’s more, data centers require cooling towers and air mechanisms to dissipate heat, which demand significant amounts of water—2 liters per kWh of energy use. A scarce and finite natural resource, AI’s water footprint even affects the local water supply, ecosystem, and biodiversity. [1]
It’s best to think that as AI technologies further advance, their environmental impact will lessen and become more manageable, and their design will eventually become more sustainable. For now, though, because newer models are typically larger, they end up compounding the cycle of increasing energy demand, and there are still significant data gaps to gauge their true short-term and long-term environmental impact. The growing integration of AI into our daily lives and apps could also lead to unconscious overuse by everyday users, making the establishment of more mindful and sustainable usage parameters essential.
How Can You Use Generative AI More Mindfully?
Making AI development, deployment, and usage more sustainable falls mostly on those who create it. However, as end users, we can set some guidelines for ourselves when it comes to leveraging these tools as effectively and mindfully as we can in our professional and personal lives.
Educate Yourself
Okay, you need to use an AI tool to get an administrative task out of the way or further streamline a process. This is a good opportunity to learn more about AI in general. As is the case with most things, educating yourself on this subject will do more good than harm. Invest some time to learn more about what it takes to develop and train an AI model. Find out as much as you can about the data it uses and how it is programmed to process it. Learning more about this topic, which will continue to remain in the spotlight until the next big thing, is an investment in being a more responsible user.
Is There Another Way To Get Your Answer?
Can you ask a colleague, a friend, or an expert about what you need to know? Can you look it up through other means, such as a web search or an online community? Yes, an internet search has its own environmental impact. Number-wise, at the end of 2023, Google’s greenhouse gas emissions (GHG) were equivalent to 14.31 million metric tons of carbon dioxide. [3] Yet, a considerable percentage of this figure was largely due to the company’s expanding data centers required for the operation and expansion of its AI services.
Research also estimated that a single prompt to ChatGPT demands five times more electricity than a web search. [1] However, with many search engines currently piloting embedding AI answers within their search results, the upcoming shifts in how we use the internet will certainly be interesting, not to mention the new numbers regarding their environmental footprint. Still, before you send a quick question to your favorite chatbot, consider whether investing some extra effort to locate the answer elsewhere is worth it.
Promote Knowledge Sharing
Recently read a quality piece on the environmental impact of AI or an unprecedented leap in its technology? Share it with your peers. If your company mandates the integration of AI tools in day-to-day workflows, set up a training session to address concerns and provide insights on the reasoning behind their implementation. Consult a professional with AI expertise to educate your employees on the intricacies of this tech and how to leverage it fully. Alternatively, put together a team that keeps track of these new trends and advancements to give the scoop to the rest of the organization. By establishing effective knowledge-sharing channels, you (and your team or company) can learn how to best utilize these emerging technologies and determine how you can get the returns you seek.
(Try To) Balance It Out
Many of us use AI because we have to, because it makes things easier, or simply because we want to explore this technological breakthrough happening before us. These are all perfectly acceptable reasons. Still, if you know you’re going to be using your go-to genAI tool with increasing frequency, try to give something back to the environment. You’ve seen the numbers and environmental effects behind AI usage. While mere individuals can’t outweigh and overcome the combined environmental impact of ginormous data centers working to keep AI running, you can still try to be more environmentally conscious. Practice sustainability, establish organization-wide programs for corporate environmental responsibility, work from home, get a bike, plant some trees, etc., to balance the scales even a tiny bit. It could be an idealistic endeavor, but with collective effort, it may eventually make a difference.
Conclusion
At this point, almost everyone has used AI in some capacity, and it’s great to give new things a try. Its innovative potential is certainly something worth exploring. What’s important is to form your own opinion on when and how you should go about it. Do you want AI to be a tutor, an assistant, a search engine, a recipe developer, or something else entirely? It’s up to you! Generative AI has the power to disrupt the world as we know it, and not just when it comes to how we do things. Its impact reaches beyond the redefinition of efficiency, encompassing almost every aspect of human activity. It all starts with our environment. Remaining mindful, seeking educational opportunities to broaden our understanding, and, of course, doing our part in ensuring a better tomorrow for future generations is more crucial than ever.
References:
[1] Explained: Generative AI’s environmental impact
[2] A Computer Scientist Breaks Down Generative AI’s Hefty Carbon Footprint
[3] Net-zero carbon