Greening AI for sustainable computing: A call for tech firms, policymakers

AI is helping reduce energy wastage and improve business process efficiencies. However, its energy use may be neutralising the overall benefits from digitisation

Technologies are providing solutions to many business and social problems. However, the solution itself is becoming a problem for climate action. Even artificial intelligence (AI)-based solutions, which are being increasingly used for climate action, are posing a new set of challenges for the world. 

 

AI is being recognised as an energy guzzler and consequently a contributor to the climate crisis. The COP29 meeting recognised the importance of AI but also flagged concerns about its impact. At COP29, UN body International Telecommunication Union (ITU) emphasised the greening of AI and computing. 

 

The energy consumption associated with AI technologies, particularly deep learning models, is substantial. Training large-scale AI models like GPT-3 and AlphaGo requires immense computational power, resulting in significant energy use. For instance, training GPT-3 required 1,287 MWh of electricity, equivalent to the annual energy consumption of over 120 US homes, according to research published by Springer.

 

Another study quoted by Springer highlighted that training a single AI model can emit as much carbon as five cars over their lifetimes. Moreover, datacenters housing AI infrastructure are major consumers of electricity.  This could lead to energy shortages caused by increasing use of AI. Gartner estimates the power required for data centres to run incremental AI-optimised servers will reach 500 terawatt-hours (TWh) per year in 2027, which is 2.6 times the level in 2023. 

 

“The explosive growth of new hyperscale data centres to implement GenAI is creating an insatiable demand for power that will exceed the ability of utility providers to expand their capacity fast enough,” said Bob Johnson, VP Analyst at Gartner. 

 

“In turn, this threatens to disrupt energy availability and lead to shortages, which will limit the growth of new data centres for GenAI and other uses from 2026.” Continuous innovation in AI and its applications is crucial for addressing global challenges like climate change. However, the computational power needed to sustain AI’s growth is doubling roughly every 100 days, leading to a significant rise in energy consumption, ITU says. The energy required to run AI tasks is also escalating, with annual growth rates between 26 per cent and 36 per cent. As organisations increasingly rely on cloud computing to support their AI initiatives, it becomes vital to establish proper guidelines and standards that mitigate the environmental impact of these technologies, according to ITU. 

 

“COP29 should propel us forward with the shared belief that we can and must reduce the environmental footprint of digital technologies while leveraging their undeniable potential to tackle the climate crisis,” said ITU Secretary-General Doreen Bogdan-Martin. “Let’s keep building our green digital momentum all the way to COP30, and with it, a more sustainable digital future for generations to come,” she said.

 

Experts believe that AI and GenAI can be made greener by a slew of steps. Tech companies will have to design algorithms in a way that they consume less energy.

 

“One of the most productive strategies in green algorithm development is the design of optimisation techniques that reduce the computational resource requirement, thus minimising energy consumption. Areas of research that are active in decreasing both the memory footprint and the computational complexity of training models include sparse training methods, quantisation techniques and energy-aware pruning and low-precision arithmetic operations,” says a paper in Science Direct. The paper further adds that selecting more computationally efficient hardware can also contribute to energy savings. Some graphic processing units (GPUs), compared to other GPUs, have substantially higher efficiency in terms of floating point operations per second (FLOPS) per watt of power usage. 

 

AI is helping reduce energy wastage and improves business process efficiencies. However, its energy use may be neutralising the overall benefits from digitisation. 

 

Technology companies and policy makers have to come together for the greening of AI now. Hopefully, the efforts to reduce the carbon footprint of AI will allow the net impact of digitisation to be climate positive.