OpenAI CEO Sam Altman recently highlighted the need for an energy breakthrough for the future of artificial intelligence, which is expected to consume much more energy than expected. Artificial intelligence has become an integral part of state-of-the-art technology, driving innovation in various fields. In the face of the climate crisis, the role of artificial intelligence is increasingly analyzed. Can AI facilitate mitigate climate change, or does its energy-intensive nature make the problem worse?
Training AI models, especially gigantic language models such as GPT-3 and GPT-4, requires significant processing power. In 2022, researchers estimated that GPT-3 training consumed at least 1,300 megawatt hours (MWh) of energy, enough to power about 130 U.S. homes for a year. As AI models become more sophisticated, their energy requirements will only escalate.
Data centers, the lifeline of artificial intelligence, have roughly the same carbon footprint as the aviation industry. According to the International Energy Agency, the estimated electricity consumption in data centers in the world in 2022 was approximately 1-1.3% of global electricity demand. Consumption is likely to double by 2026 compared to 2022. This does not include the energy used to mine cryptocurrencies, which is also significant.
AI systems also employ a lot of water for cooling purposes. Research conducted at the University of California showed that ChatGPT3 needs to drink half a liter of water for a regular conversation. GPT4 probably uses more. The study also estimated that Google’s gigantic language model, known as Lambda, used about one million liters of water for training alone.
These numbers raise concerns about the carbon footprint of AI technology. Data centers that house AI infrastructure consume huge amounts of electricity, often obtained from fossil fuels. This energy consumption contributes to greenhouse gas emissions, hampering efforts to reduce the global carbon footprint. Despite its energy needs, AI also has significant potential to combat climate change. Some technological advances can facilitate us fight climate change.
An critical issue today is whether we can develop solar panels or other materials capable of producing energy efficiently. Historically, the process has been trial and error, much like Thomas Edison tested thousands of materials to create the lightweight bulb. Today, artificial intelligence and machine learning are improving our search for fresh materials by quickly analyzing millions of potential solutions through simulation. Artificial intelligence is also helping us better understand Earth’s systems by analyzing massive data from satellites and remote sensing. This helps monitor the effects of climate change and improve responses to emergencies.
Climate crisis scientists are now using artificial intelligence to map Antarctic icebergs 10,000 times faster than humans and to track deforestation in real time to better predict weather patterns and suggest more capable waste management systems.
AI systems can optimize energy employ in homes, industries and cities, reducing overall consumption and emissions. For example, Google DeepMind artificial intelligence was used to improve the energy efficiency of Google data centers, reducing cooling costs by 40%. By analyzing historical data and forecasting future needs, the AI system optimizes refrigeration processes, significantly reducing energy consumption and greenhouse gas emissions.
Artificial intelligence was used to develop contrail forecast maps, reducing the number of contrails during test flights by 54%. Contrails are slim white lines left by airplanes and are responsible for about 35% of aviation’s contribution to global warming. They form when planes fly through humid layers, often last for hours, and trap heat in the Earth’s atmosphere. Avoiding routes that produce contrails can reduce warming.
Artificial intelligence can improve the efficiency and reliability of renewable energy sources. In the Netherlands, artificial intelligence was used to predict the output of wind farms 36 hours in advance. This project improved the integration of wind energy into the grid, ensuring a stable and sustainable energy supply by better matching production to demand.
Modern Delhi-based Blue Sky Analytics uses artificial intelligence and satellite technology to monitor air quality and analyze pollution data in real time. Their platform helps governments, businesses and individuals make informed decisions to reduce air pollution and its harmful effects on the climate and public health.
CropIn, another startup from Bengaluru, uses artificial intelligence and data analytics to support sustainable agriculture. It provides farmers with real-time insight on weather, soil conditions and crop health. This helps optimize farming operations, reduce resource waste and escalate productivity, while reducing environmental impact. CropIn’s creative approach gained recognition and funding from Google in 2023.
While AI’s energy consumption is a legitimate concern, its potential to deliver significant environmental benefits cannot be overlooked. To balance the scale, it is necessary to adopt strategies that mitigate the impact of AI on energy while maximizing its positive contribution.
Investments in energy-efficient data centers and renewable energy sources can reduce the carbon footprint of AI infrastructure. For example, Microsoft has committed to using 100% renewable energy in its data centers by 2025 and aims to achieve zero waste certification for its data centers by 2030.
Developing more energy-efficient algorithms could reduce the computing power required to train and deploy AI. AI companies are exploring ways to improve the performance of AI models using techniques such as model pruning and quantization, which preserve performance while lowering energy consumption.
Edge computing can be used to process data closer to the source, reducing the need for extensive data transmission and reducing energy consumption in centralized data centers. Retail is one industry that can greatly benefit from edge computing. By deploying high-performance AI infrastructure and applications directly in individual stores or warehouses, rather than relying on centralized corporate data centers or the cloud, retailers can significantly improve various aspects while using low energy.
By investing in equipment designed specifically for AI workloads that is energy capable and has a lower environmental impact, you can significantly reduce the carbon footprint of your AI operations. Conventional central processing units (CPUs) are not optimized for the unique requirements of AI workloads, which involve large-scale data processing and sophisticated computations.
Artificial intelligence is a double-edged sword in the fight against climate change. High energy consumption is a challenge, but the benefits are significant. Artificial intelligence can optimize energy employ, escalate the integration of renewable energy sources, reduce emissions, improve agriculture and advance climate science. By adopting sustainable practices and creative solutions, we can harness the power of artificial intelligence to fight the climate crisis and build a sustainable future.
June 5 is celebrated as World Environment Day.
(Views are personal)
Reservation
The views expressed above are those of the author.
END OF ARTICLE
