Out of sight, out of mind - thirsty, power-hungry Gen AI's hefty impact on the climate. Are we just going to ignore it?
Gen AI's ability to generate content quickly and easily is impressive, but its high energy consumption raises concerns about its environmental impact. As the tool becomes more widely used, its carbon footprint is likely to increase significantly. It is important to consider the environmental consequences of relying on such technologies for content creation. Finding ways to mitigate the environmental impact of Gen AI's operations will be crucial moving forward.
For many people, artificial intelligence has already become a part of life, helping out with repetitive, menial tasks or quick research and generating everything from job application letters to funny memes. It has also made a big positive impact in several industries including healthcare, finance, and technology, making it possible for people in these fields to innovate quickly and improve their services and offerings.
The Hidden Costs of Artificial Intelligence
Yet, as the technology advances, its hidden costs are coming into sharper focus as well. In a two-part series, CNA TODAY examines some of these covert influences and how individuals, companies, and governments may mitigate the adverse impact.
The Impact on the Environment
This week, we explore how Gen AI requires massive resources to drive its operations and growth, and how that is taking an intense toll on the environment. Last weekend, social media was seemingly spirited away into the world of Studio Ghibli.
From everyday users to governments, netizens flooded the internet with images in the distinct hand-drawn style of films such as My Neighbour Totoro and Kiki’s Delivery Services. But rather than being hand-drawn, the images inspired by the award-winning Japanese animation studio were generated using ChatGPT, a Generative Artificial Intelligence (Gen AI) tool created by OpenAI.
Environmental Impact of Gen AI
According to the International Energy Agency, a single ChatGPT query requires 10 times the amount of electricity as a Google search – 2.9 watt-hours (Wh) versus 0.3 Wh. A study by researchers at Carnegie Mellon University and machine learning platform Hugging Face in 2024 found that generating 1,000 images using several Gen AI tools requires about 2.907 kilowatt-hours (kWh) on average.
While the energy required to generate text is typically less, a study by the University of California, Riverside found that a 100-word email generated by an AI chatbot using ChatGPT’s GPT-4 model requires 0.14 kWh. That’s still enough electricity to power an LED light bulb for 14 hours.
Why Gen AI is Bad for the Environment
While traditional search engines and other traditional AI technologies also require computing power, the immense amounts needed to train and operate Gen AI models lead to an environmental impact that far outweighs that of the former, experts told CNA. Every component in the AI supply chain has its own environmental cost, said Mr Laurence Liew, director of AI Innovation at AI Singapore.
“Training a single large language model can consume more electricity than hundreds of households use annually. The training phase alone for models like GPT-4 or Claude requires thousands of specialized GPUs (Graphics Processing Units) running at full capacity for weeks or months,” he said.
Efforts to Mitigate the Environmental Impact
Several businesses and organisations in Singapore have taken proactive measures to mitigate the environmental impact of Gen AI, which they use in their operations. Singapore’s biggest bank DBS said that it has a four-lever approach to address environmental challenges, such as reducing consumption of resources and generating renewable energy.
Assoc Prof Daniel Ting, director at Singhealth's AI Office, said that the organisation is "cognizant of the amount of compute resources used" in training its AI models. Singhealth has developed several Gen AI projects like Note Buddy, which transcribes and translates doctor-patient consultations into medical notes and is used by 1,700 healthcare professionals across Singhealth.
NUHS also seeks to use renewable energy sources for compute whenever possible, said Adj Prof Ngiam of the hospital group's Gen AI usage. As for small and medium-sized enterprises (SMEs), the environmental impact of their usage of Gen AI, while not well studied, is likely minimal, said experts.
Making Gen AI Greener
While many believe that the future efficiencies of Gen AI will outweigh the current environmental costs, efforts are needed to make renewable energy more accessible so that data centers can rely on less carbon-intensive options. Companies developing AI should look at improving the efficiency of their models and hence reducing the computing power needed.
Mr Oostveen of Pure Storage said, “Apart from hardware, prioritizing smarter data management can help reduce AI’s environmental impact,” adding that large data sets and unnecessary data increase the computing power required, which in turn raises electricity and water needs.
With deliberate focus and collaborative effort across industry, government, and academia, we can ensure that AI's benefits don't come at the expense of our planet's future.