Why businesses should urgently tackle AI’s environmental impact
While there are high hopes that artificial intelligence (AI) can help tackle some of the world’s biggest environmental emergencies, there is also an environmental downside to the rapid adoption of AI and its infrastructure. Prabhakar Srinivasan from Synechron outlines why businesses should urgently tackle AI’s environmental impact.
Companies like OpenAI, Google and Microsoft, which have been instrumental in the advancement and accessibility of AI, are now raising the alarm on AI-related carbon emissions. In fact, Google’s greenhouse gas emissions in 2023 were 48% higher than in 2019, according to its latest environmental report, which highlights the increasing amounts of energy required by its data centres, exacerbated by the explosive growth of artificial intelligence.
As AI continues to develop, it’s critical that businesses and consumers both understand the environmental impact of the AI boom and work together with technology providers to establish greener approaches and solutions.
The unintended consequences of the rise of AI
While traditional AI models have had a limited carbon footprint, the rise of modern AI systems has significantly increased environmental challenges. Most large-scale AI deployments are housed in data centres, including those operated by cloud service providers – and these data centres can take a heavy toll on the planet.
Data centres not only produce significant carbon emissions but also generate substantial heat which requires energy for active cooling systems. This added demand places considerable stress on existing energy grids.
While steps have been taken to leverage sustainable power sources to reduce dependence on fossil fuels for cooling facilities, including the use of nuclear power, clean electricity technologies and renewables, we’re currently limited by the ready availability of these energy sources globally.
Addressing this environmental impact
While the global energy infrastructure plays catch-up, there are several ways to balance the scales in terms of total environmental impact. Firstly, while AI may be adding to global carbon levels, industries and sectors that have traditionally been heavy emitters, such as logistics and fleet management firms, can now adopt AI-powered route optimisation algorithms that help to reduce carbon emissions – and this is a significant step in the right direction.
Similarly, renewable energy companies can invest in AI-powered ‘digital twin’ technology. This tech enables the precise measurement of infrastructure efficiency, allowing companies to swap certain models of solar panels and assess energy efficiency and emissions without disrupting real-world installations.
Are small language models an answer?
While large language model (LLM) AI tools, like ChatGPT and Gemini, are often seen as the pioneers of the AI boom we see today, their use can be akin to taking a cannon to a knife fight; you often don’t need that level of computational power in order to write a simple shopping list or to plan an itinerary for your next trip.
Training an LLM generates approximately 284 tonnes of carbon dioxide emissions, the equivalent to 340 flights between London and New York – and the user typically utilises only a tiny portion of the model's training capabilities. So, there’s a real opportunity to shift towards the adoption of small language models (SLMs). These are faster and easier to train, while also producing superior results (given that they’re specifically trained for the purpose they’re being used for).
From an environmental perspective, these smaller models consume significantly less energy due to their dependence on less extensive server systems and smaller datasets. This greatly reduces the training burden and the associated energy costs.
In short, while some applications of AI significantly contribute to carbon emissions, others, like SLMs, can play an important role in reducing them – showcasing the need for sustainable AI deployment, and highlighting the importance of understanding and implementing these kinds of reduction strategies.
Paving the way for more sustainable AI
AI plays a critical role in addressing climate change that extends far beyond direct applications like energy optimisation – but only if it’s developed and adopted appropriately.
A recent report from the University of Sydney highlights that researchers are developing AI methods that reduce the energy required by data centres. And this innovation could actively help to reduce the carbon footprint of large language models like ChatGPT, which requires the same amount of electricity as up to 17,000 households.
Moreover, generative AI is helping to drive the creation of innovative materials that can help reduce the reliance on plastics, enhance carbon sequestration, and improve solar panel efficiency through advanced chemical properties. Scientists now have access to hundreds of AI-generated compounds to test for their physical characteristics, accelerated by progress in sustainable AI solutions.
AI is not just revolutionising everyday tasks like drafting emails or writing news articles, it’s also playing a significant impact in supercharging scientific research and discovery and leading the innovation charge.
The potential for positive change is immense, and companies that rise to the challenge of adopting this new technology in a greener way will be positioning themselves (and the planet) for greater success in the long run. By prioritising long-term thinking and strategic planning, businesses can begin to harness the power of AI to drive innovation responsibly.
