We are not talking about it, but AI will have a massive climate impact
It is rarely convenient to be climate friendly. AI fever is raging in society, with consultancies throwing around promises of increased efficiency. Universities are racing to meet the expected demand of employers for graduates with AI skills. And both private and public workplaces are experimenting with implementing generative AI in day-to-day tasks.
No matter where you look, there are high expectations for generative AI to become an inevitable part of our future. On the other hand, news coverage of generative AI’s climate footprint—with very few exceptions—is notable by its absence. But we have to address the power-hungry infrastructure that hides behind, for example, ChatGPT’s user-friendly interface.
Here it is worth repeating IDA’s communication of calculations carried out by researchers from University of Copenhagen’s Department of Computer Science: “a single ChatGPT prompt consumes [on average] 0.19 kWh or the same amount of energy as 40 recharges of your mobile phone. For the same amount of electricity, you can also drive one kilometre in your electric car or stream an hour of video.” Researchers from Cornell University predict that by 2027, AI’s demand for water could reach half of the UK’s water withdrawal.
We are in the midst of a climate crisis where even small temperature increases have fatal consequences for the ecosystems that we and future generations depend on. This is why we are surprised that politicians and decision-makers are setting ambitious targets to reduce CO2 emissions, while at the same time strategizing for a future with generative AI—a technology that drains our already depleted and limited resources.
Uncomfortable
It is now time to have that uncomfortable discussion. Because even though OpenAI’s ChatGPT may be the fastest growing consumer app, we can still create illustrations, write emails, SoMe posts, and summaries without AI. Tech giants such as Google, META, and Microsoft are working on integrating generative AI solutions into the software packages that we already use and that many workplaces and educational institutions rely on every day.
We believe that in a few years, we will be using generative AI solutions as part of this software when we search for information, write, develop ideas, program, read texts, you name it. And it will not take long for us to get used to the convenience of delegating tasks to technology to the point where we are deeply dependent on it. Not least because generative AI technology will become unavoidable once it is an integral part of our IT infrastructure and software.
This is why we insist on destroying the good and predominantly technologically optimistic atmosphere. It is now—while we still have the opportunity to say no to the technology in our software and infrastructure—that we should think about the climate.
Technology destroys ecosystems
This is where the late sociologist of science Bruno Latour’s thoughts on “the modern constitution” are important. Latour believed that we have built our Western society on a logic based on a distinction between nature and culture. That distinction has allowed society to develop at a rapid pace through a consistent and single-minded focus on human needs, resulting in an ever-increasing dependence on technology and natural resources.
Thanks to the blind logic of distinction, society has been successfully detached from nature and ecosystems. Today, we are paying the price for that through the climate crisis. Ecosystems can no longer be seen as a passive backdrop for human activities. Instead, they provide resources for our infrastructures and technologies. At the same time, our diligent use of technology has major consequences for ecosystems. This has now become clear to most people.
One might be led to believe that—in confronting generative AI’s double-edged sword of possibilities and climate footprint—we as a democracy would consider whether we can and want to accept the planetary consequences that come with it. The consequences involve further exploitation and destruction of scarce energy resources, water, and minerals. In addition, the societies that pay the price for this increased resource utilisation are rarely the societies that reap the benefits of the new technologies. This also makes it harder for us to realise the cost of our societal innovations, since we are not the ones paying most of it.
What kind of future do we want?
We therefore have to ask ourselves to what extent we, as a global society in the midst of a climate crisis, can allow ourselves to incorporate generative artificial intelligence into our workplaces, IT systems, and educational institutions. Generative AI can certainly contribute to solving critical social tasks, e.g. within research. So, it is not a question of fully rejecting generative AI—but rather of having a discussion about the role the technology should play in our society.
Researcher Kate Crawford suggests that we should create transparency around the tech giants’ massive and hidden resource and energy consumption. And that we should develop a more climate-friendly infrastructure for the development and use of generative AI. But that is not enough. We should also consider which problems we use the resource-hungry technologies to solve.
We are at a crossroads where generative AI can be integrated into our basic infrastructures—those that support our ways of learning, working, and performing mundane work tasks. But we can still choose not to.
Do we really need robots to write emails or summaries for us? Do we need avatars showing up to Teams meetings in our absence to take notes for us? Once we get used to the helping hand of generative AI, it will be hard to wean ourselves off it again. We need to discuss when—and for what type of tasks—we can justify relying on technology.
Mindless
When we talk about the use of AI today, it is often with a view to the future: AI will revolutionize work, the public sector will achieve unprecedented productivity, and both time and resources will be saved at a societal level. In fact, we need to incorporate generative AI in every area we can think of, because what if the countries we compare ourselves to are doing it while we falter? We may end up falling behind in international competition and gambling with the welfare of the future.
We agree that it is necessary to stay focused on the future, but the future-oriented mindset we advocate is one that takes a serious stance on the climate footprint of new technologies and possible negative consequences for the future. Can we, as a society, afford to mindlessly invest in generative AI?
