Did you ever wonder what would happen if any of our sci-fi movies came true?
What if humans destroy the Earth and make it uninhabitable like in WALL-E or Interstellar?
Both great movies, which I once consumed for entertainment purposes only, but now I find myself consciously thinking about them as AI becomes widely popular and accessible to the masses. I am further alarmed when there are reports which inform us that total greenhouse gas emissions keep growing year-on-year from data centers’ energy usage.
ChatGPT and Gen AI broke all records of technology adoption and entered every household in a much more tangible sense than before. Everyone is using AI now, as it should be but not without understanding the consequences and the mechanics that go behind maintaining these simple-looking functional AI tools. While I am elated with the accessibility and user-friendliness of these tools, I grow concerned about the environmental damage we cause as we use these tools in our day-to-day life. This is why it’s important that organizations and we as people incorporate Responsible and Green AI practices into our frameworks. We must act now.
As we pivot towards an AI-first reality, organizations are increasingly developing and using intelligent technology and tools to propel business reinvention. While they strive to generate business outcomes and economic value, they are also making sizable contributions to greenhouse gas emissions. Ceasing all advancement at this crucial juncture is not an option. So organizations must focus on sustainability and develop an AI eco-consciousness. Sustainability is no longer just a buzzword—it’s a fundamental requirement; and training is one of its load-bearing pillars to spread and ingrain sustainable practices.
Accenture’s Global Lead for Sustainable Technology, Sanjay Podder says “Many software developers and engineers lack the knowledge or the avenues to learn and develop expertise to be able to put Green Software Engineering into actual practice. Hence, we, at Accenture, through collaborations with the Green Software Foundation and our partners, created a set of training for our software developers and engineers so that we can champion and put Green Software Engineering in practice. To date, we have trained about 40,000 software engineers across Accenture.”
As the global lead for LearnVantage AI Academy, I believe that training plays a dual role in advancing our usage of AI and establishing sustainable practices. I understand the importance of a comprehensive, future-forward and holistic curriculum. Our general approach to skilling must include both – technical topics like Data & AI Fundamentals, Python, NLP, etc. along with Green AI Fundamentals and practices. Stage 0 is a great place to start and it’s essential to incorporate Green AI principles in the people who lead the advancement of AI for us.
This article provides a perspective on why Green AI is important and why we need to accelerate its adoption.
What is Green AI?
Green AI is about embracing energy-efficient and carbon-conscious practices to slash emissions across the entire lifecycle of AI. Picture this: by tweaking algorithms to need fewer computational resources and crafting hardware that sips (not drinks) power, Green AI could help trim down energy usage during both training and running AI models. This helps save on electricity bills while bringing down the carbon emissions from data centers and computing setups.
Why is it important?
Let’s take a look at the environmental impact of data science and data centers. Data science is the engine driving AI innovation, unlocking insights from vast troves of data to shaping decisions and accurate predictions. But behind the scenes, the energy-intensive processes raise pressing questions about sustainability. The sheer scale of data storage and processing in mammoth data centers demands hefty energy consumption for cooling and upkeep.
The data center industry, dominated by giants like Amazon AWS and Google Cloud, relies on energy-guzzling GPU chips, contributing to approximately 3.5% of global greenhouse emissions, according to the International Energy Agency.
As AI applications burgeon, so does their environmental footprint. A single large model training session can emit over 626,000 pounds of CO2—five times the lifetime emissions of an average car. Even everyday usage of large language models burns through significant carbon, posing a dire threat to global energy resources by 2040.
Maintaining infrastructure and running inference sessions also consume energy. Tools like ChatGPT and BERT, while gaining popularity, conceal hidden environmental costs that could skyrocket unless we act now.
So, how do we make generative AI green?
Green AI Solutions: Paving the Way for a Sustainable Future
Several initiatives and innovations are driving the adoption of Green AI.
For instance, the Green Software Foundation (GSF), founded by Accenture, Microsoft, GitHub, and Thoughtworks aims to reduce greenhouse gas emissions in the ICT sector by 45% by 2030. Tools like the Software Carbon Intensity technical specification help measure and manage the carbon emissions of software development processes.
Google’s environmental report also outlines several strategies the company employs to address the energy requirements of its AI advancements and researchers from Google and UC Berkeley have shown that we can slash the carbon footprint of large language models by 100 to 1,000 times with the right mix of algorithms, hardware, and energy-efficient data centers.
Apart from tools that help us measure and control emissions, there are several learning initiatives at play as well, for employees at all levels, which focus on the necessity of implementing Green AI practices and impart valuable knowledge on its efficient adoption.
While learning might seem like the underdog in discussions about Green AI solutions, its importance cannot be undermined in shaping a sustainable future. Only by training AI practitioners and integrating principles of energy efficiency, renewable energy, and responsible innovation, can we unlock the potential of AI and drive significant environmental benefits.
Green AI principles serve as the compass, ensuring data-driven initiatives maintain sustainability as a guiding star throughout their entire lifecycle. By harnessing the synergy between the two, we gain the power to catalyze transformative change, paving the way for a greener, more resilient world for future generations to inherit.