The AI paradox: Innovation at the price of sustainability?
Google’s GHG emissions have increased by 48% compared to the baseline year 2019, according to its 2024 environmental report.
This is a major setback for Google’s ambitious and exemplary goal of going carbon neutral by 2030. Google has attributed the cause of the surge in GHG emissions to increased energy consumption by data centers and supply chain emissions. This is not an isolated case as many key industry leaders have reported a similar surge in GHG emissions. The underlying reason is evident here - the booming growth of AI in the past few years.
Google’s report sheds light on the gravity of the situation. Experts are calling for better regulations and transparency in accounting for the impact of AI across industries. Let’s try to put AI’s environmental impact in perspective and explore how businesses can stay committed to their sustainability goals while retaining their competitive edge with technology.
Understanding AI’s Environmental Impact
Artificial Intelligence (AI) has become an integral part of our lives. AI’s ability to analyze vast amounts of data, identify patterns, and automate processes has revolutionized various industries. However, this rapid growth comes with environmental implications that goes beyond its massive carbon footprint.
Massive carbon footprint of large-scale AI Models
Gen-AI took the world by storm and AI has grown at a revolutionary pace ever since. As tech companies compete to create better and more powerful AI models, the process of building, operating, and maintaining AI technology requires substantial computational power, which translates to significant energy consumption and comes at a great environmental cost.
- AI is energy-intensive
The primary reason is that a significant part of electricity used for training and operating AI models still comes from fossil fuels. As a result, training AI models contributes to greenhouse gas emissions. Leaders like Google are investing in renewable energy sources but they cannot often meet the pace of mounting energy requirements for rapid large-scale tech advancements. - Application-specific effects and emissions of AI
The true impact of AI lies in how it is used and what it is used for. While training costs are significant, the day-to-day usage emissions often surpass them. For widely used models like ChatGPT, operational emissions can quickly exceed training emissions.
AI can be used in developing clean energy technologies, optimizing production processes to minimize waste and energy consumption, and predicting and negating the effects of climate change. But, AI could be also used to optimize inherently harmful activities. AI can be used for more efficient oil and gas extraction processes leading to a greater environmental footprint. - E-waste hazards generated by AI
The rapid growth of AI also results in the proliferation of e-waste. AI needs powerful hardware and infrastructure upgrades to meet the constantly evolving need to handle complex AI systems. The lifespan of a typical data center server is 3-4 years. These devices and infrastructure contain hazardous materials like lead, mercury, and cadmium that contaminate soil, water, and air. The disposal of e-waste, predicted to go over 120 million metric tonnes by 2050 by WEF raises serious concerns. This calls for responsible AI development, efficient recycling processes, and techniques to extend the lifespan of hardware. - Overreaching effects of AI: Driving consumption behavior
AI-powered automation can increase production, consumption, and waste generation. This increase in consumption behavior contributes to more emissions. For instance, AI recommendation engines and targeted ads entice consumers with personalized content. While this enhances marketing efficiency, it also fuels consumption. Fast-fashion advertising, for example, encourages frequent purchases, contributing to the fashion industry’s substantial carbon footprint which already contributes to more than 8% of global emissions. The e-commerce industry, propelled by AI-driven logistics and supply chains, has normalized rapid and frequent deliveries. This convenience comes at an environmental cost with increasing packaging waste and energy consumption.
The challenge of accounting AI emissions
The complexity of AI systems, the lack of transparency in the training and development of AI models, and the exponential rise in the use of AI across industries make it difficult to track and account for AI emissions and gauge its impact on the environment.
When accounting for emissions, it's crucial to define the boundaries. In the case of AI, should it just be development and training or also include the ongoing operation of the AI model? Quantifying Scope 3 or indirect emissions can be extremely complex in the case of AI especially when there is no standardized metric to measure AI emissions. But with the rising concerns about AI’s environmental impact, it shouldn’t be long enough to have a standardized method to track AI emissions.
How can your business use AI sustainably?
AI can be used to ensure efficient use of resources and to minimize waste. It can help implement sustainable practices by analyzing the environmental impact of a product throughout its lifecycle or optimizing production processes to minimize waste materials and energy consumption. But how do you ensure that the usage of AI in your business does not create a large carbon footprint?
- Choose energy-efficient algorithms
Prioritize AI algorithms that require less computational power to train and run. An AI model developed and trained in France can be more energy-efficient than one trained in Asia simply because of the use of renewable energy in the entire process. Choose energy-efficient AI models. - Ensure transparency and accountability
Businesses must adopt transparent practices when developing and deploying AI. Make sure that you evaluate the potential rebound effects of AI-driven efficiency gains. Use these insights in decision-making processes to ensure sustainable innovation and growth. - Monitor and optimize
Implementing standardized methods for measuring and reporting the carbon footprint of AI models will help ensure that you are on the right track. Ensure that you are continuously monitoring, reviewing, and optimizing your AI emissions. A solution like Sustanium can help you track, monitor, and report your SHG emissions based on your specific business needs. - Right-size your AI tech
Be it AI algorithms, hardware, or data, make sure that you are only using what’s enough for the task at hand. An AI model trained just for image classification generates lesser emissions than a general-purpose AI model. Choose the right algorithms, optimize your hardware, and only collect data that you need. - Choose sustainable partners
One of the major contributors to a business’s GHG emissions is the Scope 3 emissions. In Google’s report Scope 3 emissions contribute to 74% of its total emissions. Prioritizing sustainable technology partners and vendors in your business can help bring down your overall emissions when implementing AI solutions. For example, when choosing a cloud service provider, you can opt for green cloud computing providers.
The way ahead: Striking balance between AI innovation and sustainability
AI is rapidly evolving and its long-term impact on the environment is still unknown. Policymakers and governments can help minimize AI’s adverse effects by establishing standards for reporting the energy consumption and potential emissions of AI models and implementing carbon pricing schemes that incentivize the development and use of energy-efficient AI technologies. Organizations can fuel sustainable innovation by promoting research on energy-efficient algorithms and hardware, ethical AI design standards, and responsible AI implementation. Individual businesses can also play a major role by choosing sustainable practices, vendors, and AI models. And the first step is to understand your emissions.