With record-breaking heat waves globally and extreme floods affecting Europe and China, it is now a crucial moment to explore the interplay between technology and the environment, including its role as artificial intelligence (AI).
What does it take to make AI “greener”? On the one hand, we must first jointly acknowledge the tangible costs of setting up and using AI systems – which can actually be quite large. GPT-3, a recently powerful language model from OpenAI, is estimated to have consumed enough energy in training to leave a carbon footprint equivalent to driving a car from Earth to the moon and back.
There are also beneficial effects that AI can have on our relationship with the environment. A comprehensive study in 2020 assessed the potential impact of artificial intelligence on the UN’s 17 goals for sustainable development, comprehensive societal, economic and environmental achievements. The researchers found that artificial intelligence could positively enable 93 percent of environmental goals, including the creation of smart and low-carbon cities; Internet-of-Things devices and appliances capable of modulating their electricity consumption; better integration of renewable energy through intelligent networks; identification of desertification trends via satellite images; and combating marine pollution.
Cement and telecommunications
AI use in industry can serve to help the environment and reduce carbon emissions. For example, OYAK Cimento, a Turkish-based cement manufacturing group, uses artificial intelligence to significantly reduce its CO2 footprint. According to Berkan Fidan, Performance and Process Director at OYAK Cimento: “Enterprise AI-assisted process control helps increase operational efficiency, which means higher production with lower unit energy consumption. If we consider a single moderate-grade cement plant with 1 million tonnes of cement production, “Percentage of additional clinker reduction – with AI-assisted process and quality control – produces a reduction of about 7,000 tons of CO2 per year. This equates to CO2 uptake of 320,000 trees in one year.”
According to the think tank Chatham House, cement accounts for about 8 percent of CO2 emissions. Thus, there is a clear environmental need to improve the efficiency of cement production, and one tool to do so is AI.
Taking into account energy-saving constraints can drive us towards new and creative innovations in artificial intelligence.
Another example that artificial intelligence has a positive environmental impact relates to Entel, the largest Chilean telecommunications company, and sensor data to identify forest fires. Combating forest fires that have raged in many parts of the world, including Greece and Northern California, requires a concerted effort. Chile is often hit by severe climate change and catastrophic weather conditions, which previously led to the worst wildfire in Chile’s history in 2017, resulting in the burning of about 714,000 acres. For a country steeped in natural wonders, with a population and economy heavily dependent on thriving forests, any kind of wildfire is devastating.
Entel Ocean, Intel’s digital device, previously tried to identify fires using IoT sensors. These sensors act as a digital “nose” located on trees capable of detecting particles in the air. The data produced by these sensors enabled Entel Ocean to use artificial intelligence to automatically predict when a forest fire would start. “We have detected a forest fire 12 minutes before traditional methods – this is a big deal when it comes to preventing fires,” said Lenor Ferrebuz Bastidas, spokesman for enterprise digital solutions for Entel Ocean. “Considering that fire can spread in a matter of seconds, every minute helps.”
Through these applications, artificial intelligence can be a powerful tool in combating climate change. But its role also as a contributor can not be overlooked. To that end, the first step is to promote the practice of more holistic and multidimensional model evaluation. To date, the main focus of research and innovation has been to improve accuracy or create new algorithm methods. These targets often use larger and larger amounts of data, and build increasingly complex models. The most telling example is in deep learning, where computational resources increased 300,0000 times between 2012 and 2018.
Yet the relationship between model accuracy and complexity is logarithmic. For exponential increases in model size and training requirements, there are linear improvements in performance. In the pursuit of accuracy, there is less priority to develop methods with improved time-to-train or resource efficiency. Going forward, we need to recognize the trade-off between the model’s accuracy and efficiency and the model’s CO2 footprint, both during training and when drawing conclusions.
A model’s CO2 footprint can be complicated to determine and compare across modeling approaches and data center infrastructures. A reasonable place to start may be by assessing the number of floating point operations – that is, a discrete count of how many simple mathematical operations (eg multiplication, division, addition, subtraction, and variable assignment) – to train a model. . This factor and others may affect the energy consumption of the model architecture and the training resources such as hardware such as GPUs or CPUs. In addition, the physical considerations of storing and cooling the servers come into play. As a final complication, it also matters where the energy comes from. Energy primarily from renewable resources compared to natural gas or coal will have a reduced CO2 footprint.
Let us ask, “How much more can we do with less?” Taking into account energy-saving constraints can drive us towards new and creative innovations in artificial intelligence. By turning to this mindset instead of bigger is always better, and by pursuing AI applications in the environmental field, AI can remain at the forefront and become a sustainable technology of the future and an important asset in protecting our global climate.