By Maxime Vermeir, Senior Director of AI Strategy at ABBYY
Sustainability is becoming both a defining challenge and opportunity for Artificial Intelligence (AI). As we look towards 2025, businesses will increasingly need to find ways to balance innovation with the need to be environmentally responsible.
Generative AI’s impact on the environment cannot be overlooked. Large language models (LLMs) in particular have high energy demands, with global AI energy demand projected to exceed the annual electricity consumption of a country the size of Belgium by 2026.
Companies must navigate massive stores of data which use huge amounts of energy. Not only does this raise risks in the way of ethics, accuracy, and privacy, but it also exacerbates the amount of energy required to use the tools. To give an idea of the difference in electricity demand, a Google search consumes 0.3 watt-hours of electricity, while OpenAI’s ChatGPT uses 2.9 watt-hours, nearly 10 times more energy.
Many of these companies currently rely on carbon offsetting, but this approach is likely to remain effective only in the short to medium term as industries work toward sustainable solutions to decrease AI’s reliance on fossil fuels.
AI could be the solution to its own problem
While the sustainability issues with Generative AI is a big potential problem, the technology also has potential to reduce its own emissions. A study by the Boston Consulting Group found that AI could help mitigate 5 to 10% of GHG emissions by 2030, if it is “used wisely”.
AI can be leveraged to optimise the performance and operation of wind turbines, solar panels, electric vehicles, and energy storage systems like batteries. One example is through analysing large amounts of data in real-time, AI can enhance energy output, predict maintenance needs, and reduce downtime for renewable energy systems. It can also improve energy efficiency through automating processes and optimizing energy consumption patterns. By minimising unnecessary processes, AI can be used to help reduce overall consumption.
Purpose-built AI can keep emissions down
One key way that businesses can keep emissions from AI down is through turning to purpose-built AI models. Balancing innovation with sustainability requires optimising the energy efficiency of AI models and intelligently managing renewable resources. Developing purpose-built AI models offers a promising solution to significantly reduce energy consumption while maintaining performance.
Businesses investing in general-purpose AI tools may find that these advanced models are sometimes applied to tasks too simple for their level of sophistication, which can result in unnecessary energy intensity. To address this, adopting more efficient, task-specific AI models can significantly reduce energy consumption while maintaining effectiveness.
Switching to purpose-built AI such as small language models (SLMs) can significantly reduce energy consumption. These solutions are built for specific tasks and tailored to improve accuracy in real-world scenarios. At ABBYY, for example, we train our machine learning and natural language processing (NLP) models to read and understand documents that run through enterprise systems just like a human. With pre-trained AI skills to process highly specific document types with 95% accuracy, organizations can save trees by eliminating the use of paper while also reducing the amount of carbon emitted through cumbersome document management processes.
Digital twins may be key to sustainability
One example of purpose-built AI that can be particularly helpful for reducing the environmental impact of high-emission sectors like energy, materials, and mobility, is digital twins.
Digital twin technology is already transformative for businesses in reducing emissions and optimising resource use, and most organizations (57%) believe digital twin technology is critical to improving sustainability efforts.
A digital twin is a virtual model of a physical object or system. It spans the object’s full lifecycle and is updated using real-time data to simulate its behaviour. Its primary purpose is to remotely monitor an object’s performance, enabling businesses to identify potential problems and make informed decisions to enhance the original physical asset.
The cost of implementing a digital twin can vary widely, ranging from $65,000 to over $1 million, depending on the complexity and scale. Simpler digital twins might involve basic modelling and data integration, while more advanced versions incorporate real-time analytics and machine learning. Despite the upfront investment, the long-term benefits can make digital twins an increasingly popular solution for organisations in meeting sustainability goals.
Regulation can help keep businesses accountable
New technologies like digital twins can be transformative, but it is important to keep businesses accountable for their emissions even as they explore ways to reduce them through the use of AI. There’s currently only a minimal legal framework governing AI, and legislation has largely overlooked its environmental impact, focusing instead on privacy and other ethical concerns.
The EU’s AI Act, for example, seeks to regulate AI systems by categorising them based on their risk levels. It mandates transparency and safety in AI development, potentially fostering more responsible usage. However, its effectiveness will largely depend on how well the provisions are enforced.
There is a pressing need for clearer national, regional, and international regulations addressing AI’s energy consumption, particularly given the energy sector’s vital role in the global economy and its significance for climate objectives.
While regulation is crucial for ensuring responsible AI practices, businesses must also take an active role. Beyond merely complying with rules, they must hold themselves accountable for the environmental consequences of their AI usage. It’s up to business leaders to take the initiative and ensure transparency and accountability for sustainability credentials are upheld when implementing AI.
I believe the path forward involves combining sustainability with operational intelligence. By integrating renewable energy management and optimizing resource allocation, businesses can minimize their environmental impact without compromising innovation.
Businesses need to address the environmental implications of AI now. Companies willing to embrace the challenge of managing their AI emissions effectively can not only gain substantial economic advantages but also establish a new standard for a more sustainable future.
The convergence of tech and brands 