12 hours ago
AI and the environment
The writer is an academic and researcher. He is also the author of Development, Poverty, and Power in Pakistan, available from Routledge
Listen to article
For Gen X people like me, who are trying to get used to the new world of AI, like we learnt using the computer, and then the Internet many years ago, it is intriguing to see how AI is becoming integrated into our lives. For researchers like me, AI is making it easier to navigate Internet searches, and to synthesise relevant literature. Besides such novice applications of AI, however, this evolving technology is going to start playing an increasingly prominent role in more salient aspects of our lives ranging from healthcare, education, manufacturing, agriculture, and even warfare.
There are also legitimate reasons to be wary of AI's power. AI is making it much easier to spread disinformation, enable fraud, and cause conflicts to become deadlier. Moreover, AI, like many other technologies that we have become so dependent on in our consumerist world, ranging from cars to cell phone, has significant environmental impacts. This heavy ecological footprint of AI is more concerning to me than speculations about AI dominating or replacing humans.
AI has a much larger environmental impact than many of the other innovations we now depend on, due to the exorbitant amount of energy needed to operate and train AI systems, and because of the e-waste produced by the hardware used to run AI. Training and operationalising large language models such as ChatGPT depend on energy still being generated via fossil fuels, which is leading to more carbon emissions, and increased global warming. Each ChatGPT question is estimated to use around 10 times more electricity than a traditional Google search.
Producing and disposing of AI hardware also generates a lot of e-waste comprised of harmful chemicals. Running AI models need a lot of water too, to cool the data centres which house massive servers, and to cool thermoelectric or hydroelectric plants which supply electricity for these data centers. The race to produce AI is also compelling major tech giants to walk back on their earlier environmental pledges. Consider, for instance, the case of Google. A few years ago, Google set an ambitious target to address climate change by becoming 'net zero' emissions, but now the company's emissions are growing due to Goggle's bid to become a leader in AI.
As the AI industry continues to grow, its environmental impact will grow too. However, as is the case of ecological destruction caused by over consumption of other products, the environmental impacts of AI will not be evenly distributed across different regions or socio-economic classes. The benefits of AI will not be evenly spread either. Higher income countries are better poised to capture economic value from AI because they already have better digital infrastructure, more AI development resources, and advanced data systems.
Better off households will be able to enjoy the benefits of AI, while having more resilience in terms of shielding themselves from its adverse impacts. Conversely, the quest to produce more AI may cause exploitation in poorer countries that provide the critical resources needed for AI. This is not a speculative statement, but one based on ground realities.
Consider, for instance, the dismal condition of miners, including children, in poor African countries like Congo, who are toiling away to produce cobalt to power batteries used to run electric cars, and our phones. Al will require many more of these critical resources, potentially leading to even more exploitation of people and natural environments in resource-rich but poor countries.
It is important to improve the energy efficiency of AI models and data centers, and to use renewable energy sources to power AI data centres. Moreover, it is also vital to promote more sustainable mining and manufacturing practices and improve e-waste management to reduce the amount of harmful chemicals entering the environment. However, whether these efforts will be paid more attention than maximising profits within this highly unregulated new domain of human innovation remains to be seen.
