Green Intelligence: Why Data And AI Must Become More Sustainable

    Published on:

    As big data, machine learning and artificial intelligence continue to dominate information technology, experts express concern about the environmental costs of computation, primarily the carbon and greenhouse gas emissions of data and AI. .

    The problem shows no signs of slowing down. As a result of the COVID-19 pandemic, the deployment of data and AI has increased exponentially as the demand for digital transformation increases.

    MIT report The cloud’s carbon footprint is currently larger than the entire airline industry, and a single data center can consume as much electricity as 50,000 homes.

    On the other hand, the datasets used to train AI are getting larger and require enormous amounts of energy to run. MIT Technology Review reports that just training one AI model can emit over 626,00 pounds of carbon equivalent. That’s almost five times the lifetime emissions of the average American car.

    See why it’s important for businesses to address how data storage and AI contribute to greenhouse gas emissions, and what we can do to mitigate the impact of this ongoing problem. Let’s look at.

    Why We Should Address This Problem

    Sanjay Podder, Managing Director and Global Leader in Technology Sustainability Innovation at Accenture, said: To tell The exponential growth of data and the accompanying increase in energy demand could actually impede and hinder global progress against climate change.

    Currently, the AI ​​community takes a “bigger is better” attitude when it comes to data and artificial intelligence, but that approach threatens to lead to major environmental disruptions in the future.

    Tech professionals have to expend more and more energy to build ever-larger models, and performance gains go down.

    For example, the AI ​​underlying a self-driving car must be trained to learn to drive. Once the initial training is complete, the AI ​​model of the self-driving car performs continuous inferences to help it navigate its environment. This process happens every day as long as you use your vehicle. This requires a lot of energy for just one car.

    Bold and thoughtful initiatives are needed to put the AI ​​field on a more sustainable trajectory.

    Suggestions for addressing the sustainability impact of AI

    What can businesses do to drive innovation while reducing the environmental impact of AI and big data? Below are some suggestions for data sustainability.

    Consider how your environmental impact is measured. Carbon accounting needs to be improved by providing faster and more accurate data on carbon footprint and sustainability impact. Tools like Salesforce’s Net Zero Cloud, SustainLife, and Microsoft Cloud for Sustainability help companies visualize and understand failures and find opportunities for improvement.

    Estimate the carbon footprint of an AI model. of Machine learning emissions calculator It helps practitioners perform estimations based on factors such as cloud provider, geographic region, and hardware.

    Find out how and where your data is stored. Some of the biggest jobs in machine learning could be moved to more carbon-friendly regions of the world. For example, Montreal, Canada has a number of data centers powered by hydroelectric power.

    Improves clarity and measurement. AI researchers should include measurements of the amount of energy emitted by the model alongside performance and accuracy metrics when publishing results for new models.

    Follow Google’s “4M” best practices. Google has identified four best practices known as the “4Ms”. This can significantly reduce the energy and carbon footprint of everyone using Google Cloud services. These include choosing efficient machine learning model architectures, using optimized processors and systems for ML training, computing in the cloud instead of on-premises, and a map for choosing where to have the cleanest energy. includes optimizations for By following these practices, Google claims 100 times less energy and 1,000 times less emissions.

    How to work towards a new AI paradigm

    As the adoption of AI and machine learning techniques into society continues to accelerate, we must consider how these tools and systems are impacting our environment. Unless there is a willingness to reform the AI ​​research agenda today and increase transparency on this issue, the world of AI could end up pushing us in the fight to slow climate change.


    Leave a Reply

    Please enter your comment!
    Please enter your name here