- While AI has the potential to be a significant tool in combating climate change, its position as a source of emissions must not be disregarded.
- The first step toward greener AI is to encourage more comprehensive and multifaceted model assessment.
- AI may be a significant tool in the battle against changing climate if we alter our mentality that larger is always good and pursue AI use instances in the environmental arena.
With record heat waves throughout the world and catastrophic flooding affecting China and Europe, now is a critical time to examine the relationship between the environment and technology, particularly the impact of AI.
What would it need to make AI more environmentally friendly? On the one hand, we must all acknowledge that there are real costs to developing and deploying AI systems — prices that can be extremely high. GPT-3, a recent strong language model developed by OpenAI, is predicted to have used enough power in training to create a carbon footprint comparable to transporting a car from Earth to the moon and back.
AI can also have a positive influence on our connection with the environment. In 2020, thorough research examined the possible influence of AI on the 17 Sustainable Development Goals of the United Nations, covering societal, economic, and environmental implications. The researchers discovered that AI could positively enable 93 percent of the environmental goals, including the development of Internet-of-Things devices, smart and low-carbon cities and appliances that can regulate electricity consumption, better incorporation of sustainable power through smart grids, the recognition of desertification trends using satellite imagery, and overcoming marine pollution.
Telecom And Cement
AI application cases in business can benefit the environment by lowering carbon emissions. OYAK Cimento, a Turkish cement production company, is utilizing AI to drastically decrease its carbon impact. Berkan Fidan, Performance & Process Director of OYAK Cimento, says: “Enterprise AI-assisted process control helps to increase operational efficiency, which means higher production with lower unit energy consumption. If we consider a single moderate capacity level cement plant with 1 million tons of cement production, just a 1% of additional clinker reduction – with AI-assisted process and quality control – produces a reduction of around 7,000 tons of CO2 per year. This equals CO2 absorption of 320,000 trees in a year.”
According to the research group Chatham House, cement contributes around 8% of emissions of CO2. As a result, there is an obvious environmental necessity to increase efficiency in cement manufacture, and AI is one instrument for doing so.
An example of AI having a beneficial environmental impact is Entel, Chile’s leading telecom provider, and the use of sensor data to detect forest fires. Fighting forest fires that have raged in many places of the world, including Greece and Northern California, requires a collective effort. Chile is regularly hit by severe climate change and disastrous weather conditions, which previously culminated in the largest wildfire in Chile’s history, which burned over 714,000 acres in 2017. Any sort of wildfire is a tragic catastrophe in a country rich in natural splendor, with a people and economy that rely greatly on flourishing woods.
Entel Ocean, the company’s digital division, attempted to detect fires sooner by utilizing IoT sensors. These sensors, which are mounted on trees, function as a digital “nose” capable of detecting particles from the air. Entel Ocean was able to utilize the data generated by these sensors to apply AI to forecast when a wildfire would occur. “We have been detecting a forest fire 12 minutes before traditional methods – this is a big deal when it comes to preventing fires,” Lenor Ferrebuz Bastidas, Entel Ocean’s business digital solutions spokesman, states “Considering fire can spread in a matter of seconds, every minute helps.”
Trade-Offs
AI may be a significant aid in combating climate change through these applications. However, its importance as a contributor should not be ignored. To that aim, the initial step is to encourage the use of more comprehensive and multifaceted model evaluation techniques. Until now, the primary focus of research and innovation has been on increasing accuracy or developing novel algorithm techniques. These goals frequently necessitate the collection of ever-increasing quantities of data in order to construct ever-more sophisticated models. The most egregious example is deep learning, where computational resources increased 300,0000 times between 2012 and 2018.
The connection between model complexity and accuracy, however, is logarithmic. There are linear gains in performance for exponential increases in model size and training needs. In the pursuit of accuracy, less emphasis is placed on creating approaches that increase time-to-train or optimizing resources. Moving forward, we must acknowledge the trade-off among both model efficiency and accuracy, as well as the model’s carbon footprint, both during inference and training.
A model’s carbon footprint might be difficult to calculate and compare across modeling techniques and data center infrastructures. A good place to start is to count the number of floating-point operations required to train a model, which is a discrete count of how many simple mathematical operations (such as multiplication, division, addition, subtraction, and variable assignment) must be done. This aspect, as well as others, can have an influence on energy usage, as can the model’s architecture and training resources, such as GPUs or CPUs.Furthermore, practical concerns like server storage well as cooling come into the equation. Finally, it is important to know where the energy is coming from. Energy derived largely from renewable energy sources will have a lower carbon footprint than coal or natural gas.
Let’s ask: “How much more can we do with less?” Considering energy-saving restrictions may push us to new and innovative AI breakthroughs. By shifting away from the attitude that larger is always good and pursuing AI use instances in the environmental sector, AI can stay on the cutting edge, becoming a future sustainable technology and a key asset in the preservation of our planet’s climate.
