Will AI be good for the environment? 

When generative AI technology suddenly exploded into the mainstream towards the end of 2022, it was immediately clear that this new technology was going to change everything. What wasn’t so clear was whether or not this would be a good thing, and what the side effects of this rapid change might be. 

As a longtime advocate of digital sustainability and author of the book Sustainable Web Design, I was immediately aware that this new technology would require a lot of energy, but also that it could be have effects on human behaviour and evolution that impact the environment in unexpected ways. 

I began the new year of 2023 by publishing an interview that I conducted with ChatGPT about the social and environmental impacts of artificial intelligence and I was surprised to find that it openly acknowledged the risks and didn’t just hype the potential benefits. In the months that followed throughout this past year, we’ve included several updates in our Curiously Green newsletter about new research into the environmental impacts of AI and I have to admit that I have found it concerning. 

However, I’ve also been aware that this new technology brings with it an unusually high degree of uncertainty and while that can be unsettling, it brings as many possibilities as it does threats. I’ve tried to keep an open mind and observe the bigger picture as to how the evolution of AI might play out and my willingness to listen to both sides led me to be invited to facilitate a debate for Innovate UK KTN on whether AI is good for environmental sustainability. 

The debate included four experts on the subject of AI sustainability, with Chris Lloyd-Jones and Dawn Geatches arguing that AI is not good for the environment, and Joyeeta Das and Maria Perez Ortiz arguing that it is good for the environment. It was a truly fascinating debate with diverse perspectives and the one hour long video is now available to watch online. 

In this post, I’m going to try to summarise the main points put forward by the four speakers and share some of my own thoughts on AI sustainability at this point in time. And seeing as it’s panto season, you’ll have to forgive my choice of headings! 

So, is AI good for environmental sustainability?

Chris Lloyd-Jones kicked off the debate by explaining just how computationally intensive AI is, both in its training and operation. He compared this to doing similar tasks without AI, questioning whether we really need to embed AI into everything when more efficient, simpler technologies are more than capable of doing many jobs, but he also went further and compared it to the energy efficiency of the human brain. Apparently, training a large deep learning model uses roughly the amount of energy used by a human brain over its entire lifetime (about 78 years). And that’s not factoring in the huge amounts of energy then used to process all of the queries sent to the AI after it has been trained. 

Chris highlighted that while there is always progress being made in improving the energy efficiency of digital hardware and software, the exponential growth in our usage that AI is driving will stand in the way of actually decarbonising the digital sector.

Dawn Geatches, speaking on behalf of herself and her colleague Caroline Chibelushi, then built on Chris’s argument by highlighting the very physical nature of the data centres that sit behind these new AI technologies. This physical infrastructure has a significant impact on the environment, not just in terms of energy consumption and greenhouse gas emissions but also the water used for cooling servers, much of it from potable sources. Add to this the impact of mining the materials to manufacture the huge number of servers required, the electronic waste that is then generated by the high turnover of these servers as well as the land used to construct the data centres and you have a trail of environmental destruction. 

She highlighted the world’s largest data centre, Switch’s Citadel campus near Reno in Nevada, which when complete will occupy 7.3 million square feet, equivalent to roughly 80 football pitches. If it were to be powered entirely by solar energy, the solar arrays required to generate its 650 megawatts of power would then cover an additional 3250 football pitches. 

All of this builds a compelling and somewhat disturbing picture that in Chris’s words, these new AI technologies are “fundamentally unsustainable”.

But not everyone agrees. And that brings us on to the second half of the debate arguing that AI is good for the environment. 

Joyeeta Das opened the second half of the debate by acknowledging that AI does use a lot of physical resources and that this is a legitimate concern. However, she countered the previous arguments by highlighting that AI technology could deliver exponential returns on the energy and resources used, helping us to decarbonize society more rapidly and solve other environmental issues. For example, if it uses some additional energy but can help us change our energy systems and societal consumption patterns, then on balance there could be a significant environmental benefit.

She acknowledged that AI could of course be used in the wrong way and drive increased consumerism, but emphasised that this is even more reason to embrace AI for environmental progress. If environmentalists sit on the side lines and boycott AI because of its environmental impact then we would miss out on the benefits and still suffer the downsides of widespread adoption. It was a compelling case for why the environmental movement needs to embrace AI to accelerate positive change, even if it is a case of fighting fire with fire. 

This was followed by Maria Perez-Ortiz, who argued that AI technology has huge potential to help us understand environmental challenges like climate change better, as well as to develop and implement more effective solutions such as integrating renewable energy sources more smoothly into power grids. The biggest opportunity though, might well be in the field of scientific discovery, where AI could accelerate progress and lead us to new, currently unseen solutions to biological and environmental problems. If scientific progress is what we need to solve the environmental crisis that we face, then AI could be the most powerful tool we’ve ever had to accelerate that progress.

Maria highlighted some good reasons to be optimistic about AI but she also acknowledged some risks, such as the accelerated spread of disinformation and the potential to increase inequality as only the more privileged groups in society get carried upward by the technology they have access to. She wrapped up by stating that while there is huge potential for AI to benefit the environment, the downsides are very much real and the benefits are currently mostly hypothetical.

I think Maria’s conclusion summed up the debate nicely. There are very real negative environmental impacts of AI that were acknowledged by all four speakers but there are also some exciting opportunities that could benefit the environment. However, these benefits are largely unproven and are arguably more dependent on how we humans decide to apply the technology than on the technology itself. The question is, can we be trusted to use it wisely? The jury is out.

Then again, maybe we aren’t as important as we think we are. In the Q&A after the debate, the question was asked as to whether AI would simply solve its own problems. It’s a great question, as if it develops the capabilities that are currently being hyped then this would surely be the case.

Maybe this could happen the future, but there are no guarantees and history teaches us that we should be careful not to always believe the hype. In response to this question, Chris emphasised that we need to direct the development and usage of AI and can’t simply take a passive approach by relying on automated solutions.

My own experience confirms this. I recently worked with generative AI tools to co-create short stories of hope about how humanity can overcome the big challenges it faces, in a series called Humanitas et Machina. In this process, it became clear to me that current mainstream AI tools still need a lot of guidance and are a long way off being able to solve complex environmental and social challenges on their own. Hopefully this will change with time and AI will gradually become a force that automatically helps guide humanity to a positive future, but at this time it is little more than wishful thinking. 

The debate covered a lot of important perspectives and provided a lot of food for thought, but I think there was one very important thing missing. If we are asking whether AI is good for environmental sustainability, then we surely need to consider the bigger picture of what AI development could really mean for humanity if we succeed in creating something more intelligent than ourselves. The truth is that there is no way of knowing, but there are a growing number of AI experts that are deeply concerned that we might be about to let the genie out of the bottle.

Geoffrey Hinton, known as the Godfather of AI, recently stated “I’m in the camp that thinks this is an existential risk, and it’s close enough that we ought to be working very hard right now and putting a lot of resources into figuring out what we can do about it…I think it’s quite conceivable that humanity is just a passing phase in the evolution of intelligence”.

Whatever the truth turns out to be, it’s clear to me at this stage that we shouldn’t be complacent in assuming that AI will be the solution to our problems.

If we are talking about environmental sustainability, then surely an existential risk should be the primary focus of our attention. Shouldn’t it? On the other hand, it’s clear that there are some among the tech elite such as Google co-founder Larry Page who don’t think so, and who feel that humanity being replaced by a greater intelligence would represent evolutionary progress. Likewise, there are even some in the environmental movement who think that the elimination of the human species by a super intelligence might be exactly what the Earth needs in order to recover and flourish. 

Personally, while I can see the cold logic of these perspectives, I find them chillingly dystopian. I believe that human life has fundamental value in the universe and that we should not be so hasty in ushering in our replacement. For me, this debate about AI and environmental sustainability is about more than just how much energy this technology might consume. It raises some big questions about who and what we are as humans, our relationship with our environment and what type of future we want. AI might offer us some incredible opportunities, but the journey ahead is a perilous one that we should venture into with a healthy degree of caution.