by Dr Azhar Yousuf
Given that AI systems rely heavily on electricity and vast natural resources, from both cost and environmental perspectives, there is little justification for being courteous to machines. This seemingly trivial cultural shift accentuates a larger issue – the detrimental impact of AI on the environment when deployed on a grand scale.
Artificial Intelligence (AI) is widely regarded as a transformative technology with the potential to revolutionise every aspect of human life, from personal interactions to professional environments and global challenges. Yet, despite its promise, AI is not without its drawbacks. One of the most pressing concerns lies in its substantial demand for computing power and energy, particularly during the training and inference phases of AI models. This extensive energy consumption translates into a considerable environmental cost, as the proliferation of colossal data centres, which house AI servers, requires vast amounts of electricity and other natural resources, thereby exacerbating greenhouse gas emissions.
The environmental toll of AI extends even to seemingly trivial tasks. Simple AI operations, such as responding to a query on ChatGPT, can consume disproportionately high amounts of energy – up to thirty times more than a standard web search, according to research findings. This phenomenon underscores the significant energy demands embedded in everyday interactions with AI. For instance, users inadvertently contribute to heightened energy consumption by including superfluous words like “Please” or “Thank you” in their prompts. While these courtesies may appear inconsequential, they impose additional computational work, compelling AI systems to process extra words that hold no substantive value, thereby squandering energy.
The implications of this energy expenditure extend beyond the environmental sphere to the financial domain. Sam Altman, the chief executive of OpenAI, recently highlighted the financial burden imposed by unnecessary politeness in AI interactions. Altman suggested that refraining from pleasantries in AI prompts could mitigate some of the substantial financial losses incurred by tech giants, as every additional word triggers computational work that incurs both energy costs and financial outlays. The cumulative effect of such seemingly negligible actions underscores the broader impact of AI’s energy-intensive infrastructure on both the environment and the economy.
In human-to-human conversations, politeness is not only a social expectation but also a virtue ingrained in cultural norms. However, the same principle does not extend to interactions with artificial intelligence. Given that AI systems rely heavily on electricity and vast natural resources, from both cost and environmental perspectives, there is little justification for being courteous to machines. This seemingly trivial cultural shift underscores a larger issue – the detrimental impact of AI on the environment when deployed on a grand scale.
The rise of AI is poised to exert significant pressure on the energy sector. Projections indicate that current energy consumption could double by 2030, exacerbating the carbon footprint of AI systems. Meanwhile, data centres housing AI servers produce substantial electronic waste laden with hazardous materials such as mercury and lead. Furthermore, they consume vast quantities of water to cool their systems, compounding the environmental toll.
Consider the recent remark by OpenAI’s chief executive, who quipped, “Our GPUs are melting,” during the Studio Ghibli social media frenzy. This statement reflects a deeper crisis. AI infrastructure is projected to consume six times more water than Denmark, a country with a population of around six million. This is particularly alarming given that nearly a quarter of the global population already lacks access to clean water.
As the AI boom accelerates, a pressing question emerges: Are we sacrificing the planet to sustain digital brains? Unlike the gold rush, where selling shovels to miners proved profitable, the unchecked expansion of AI may not be a wise choice without implementing sustainable practices. The urgency lies in developing more energy-efficient AI systems, lest the technology that promises to address climate change paradoxically becomes its greatest contributor.
(The author is a Postdoctoral Fellow in the Department of Computer Science and Engineering at the Indian Institute of Technology Bombay (IIT Bombay). Ideas are personal.)
