Can AI solve its sustainability problem?

Find the Latest SoC Solutions
for... Automotive... IoT... Security... Audio... Video

Increasing workloads and the growth of HPC and AI systems are creating a carbon crisis for research, writes Eugenia Bahit

www.scientific-computing.com/, May. 15, 2023 – 

With the proliferation of artificial intelligence (AI) and machine learning (ML), the high-performance computing industry's increasing workloads are contributing to environmental damage. It may be confusing that cutting-edge technology such as AI, which helps humanity in so many fields, can have a negative impact and put an entire industry in check. Still, the increase in AI workloads is becoming a problem that needs urgent attention.

What causes increasing AI/ML workloads?

Artificial intelligence is advancing fast, and workloads are becoming bigger due to data volume and algorithm complexity. Developing an intelligent machine requires the ability to make inferences that involve applying logical rules to deduce a possible conclusion. Current ML models do this by storing and classifying astronomical volumes of data, which is one of the factors responsible for large workloads.

The complexity of deep learning models – used to understand the data and apply the logical rules – typically requires recursion to process millions of parameters, demanding lots of resources. Therefore, the more complex the algorithms, the greater the workload.

Optimising the ML workflows to reduce workloads is not easy. According to Dion Harris[1], Head of Product Marketing for Accelerated Computing in HPC, AI, and Quantum Computing at Nvidia, some HPC users may not have the specific skills to optimise AI/ML workflows, and AI/ML experts may not have the experience to optimise them on HPC systems. Harris sees a gap between researchers with applied AI backgrounds and those with scientific computing knowledge.

click here to read more...

 Back