The Energy Behind AI

Generative AI Leaves a Bigger Carbon Footprint Than You Think

By Novadella Zen

We’re discussing it again. Breakthroughs in technology just can’t seem to exist without being connected to environmental issues.

I remember seeing a post online about ‘deleting old emails to save trees’.

Turns out it wasn’t lying. Worse yet - there’s actually more to it than just emails.

            When the first craze of ChatGPT happened in 2022, users across the world were introduced to conventional forms of generative AI. Today, usage of these tools have spiked to millions, if not billions of times every day. We’re talking chatbots (ChatGPT, Gemini, Aria), image generators (Dall-E, Midjourney), and AI assistants integrated on popular apps and services (Grammarly, Canva, WhatsApp). Lesser available to the public but key to a lot of emerging companies include search engine optimization, predictive analytics … you name it.

            Data centres are the heart of the internet, and there’s energy needed to maintain them. Something as simple as one search on Google takes up energy to output. The way Google’s search engine processes search queries is only a matter of looking up existing databases, or in this case, web pages. Now, generative software would have to conjure more specific and unique responses, thus demanding nearly 10 times the energy usage. However, the core of this issue doesn’t lie within the products themselves, but rather the processes required to keep them running.

            Every AI model requires datasets to be ‘fed’ and ‘trained’ with. That’s where they get the raw materials to piece together new information. Where is data placed in? Data centres. Every bit of information from every corner of the internet: text, images, video – It’s stunning to hear about the sheer amount of resources required to run these structures. Electricity to power supercomputers, water to cool CPUs, tonnes of carbon dioxide gases emitted daily. We have also yet to account for hardware and equipment used up to construct them from the bottom up.

            This might sound old, but it’s about time we give back to nature. Simply avoiding the use of generative AI platforms for personal use is a good first step in the direction of prevention; but this is yet another impending task to tackle in the gigantic picture of ‘saving the earth’. As unfortunate as it sounds, it can’t be achieved by eliminating one contributing factor. Thankfully, a large amount of responsibility resides within larger companies namely OpenAI and Google. Reportedly, they have begun the initiative to optimize energy efficiency in the future. For now, we can expect more companies to make wiser decisions about the environmental impacts of their products in the future.

 

What would you do to ensure a sustainable future with AI in the picture?