Generative AI, whether in the form of images created from prompts or large language models (LLMs) that process and understand text, such as the highly popular ChatGPT platform by OpenAI, has taken the world by storm. It offers an easy way for individuals, who otherwise wouldn’t have access to powerful computing resources, to utilise advanced tools. However, while someone at home may simply type in a whimsical request for an image or pose a trivial question they could have answered themselves, the true cost of offloading these tasks to AI is often overlooked.
Increasingly, environmentalists are sounding the alarm about how the rise of AI has led to a surge in emissions, which, if unaddressed, could become a significant contributor to climate change. Although we might be using these services on personal devices, and it may seem no different from how we used to operate computers, the reality is that these queries are processed through vast data centres. These facilities require enormous amounts of electricity to keep running and cool around the clock for AI operations. Even the most basic request for an image of a simple object can lead to considerable energy consumption in these data centres.
As of 2022, the combined electricity usage of the AI and cryptocurrency sectors, which rely on similar principles, was nearly 460 terawatt hours—amounting to 2% of total global electricity production. While this may seem small, the rapid growth of these technologies, alongside their still limited practical applications, raises concerns. We must consider whether the benefits of a single query to an LLM or a generative image outweigh the economic and environmental costs of fulfilling that demand.
AI also presents numerous ethical issues, such as the use of work created by other artists and writers. Now, with the addition of this environmental perspective, it’s clear that the future of AI requires careful thought and responsible action.