AI isn’t nearly constructing smarter machines. It’s about constructing a greener world.
From optimizing power use to decreasing emissions, AI and accelerated computing are serving to industries deal with a number of the world’s hardest environmental challenges.
As Joshua Parker, senior director of company sustainability at NVIDIA, explains within the newest episode of NVIDIA’s AI Podcast, these applied sciences are powering a brand new period of power effectivity.
Parker, a seasoned sustainability skilled with a background in legislation and engineering, has led sustainability initiatives at main firms like Western Digital, the place he developed company sustainability methods. His experience spans company sustainability, mental property and environmental affect.
AI can, in reality, assist scale back power consumption. And it’s doing it in some stunning methods.
AI methods themselves use power, after all, however the large story is how AI and accelerated computing are serving to different methods save power. Take knowledge facilities, as an example.
They’re the spine of AI, housing the highly effective methods that crunch the info wanted for AI to work in providers like chatbots, AI-powered search and content material era. Globally, knowledge facilities account for about 2% of whole power consumption, and AI-specific facilities characterize solely a tiny fraction of that.
“AI nonetheless accounts for a tiny, tiny fraction of general power consumption globally,” Parker mentioned. But, the potential for AI to optimize power use is huge.
Regardless of this, AI’s actual superpower lies in its potential to optimize.
How? By utilizing accelerated computing platforms that mix GPUs and CPUs. GPUs are designed to deal with complicated computations rapidly and effectively.
Actually, these methods might be as much as 20x extra energy-efficient than conventional CPU-only methods for AI inference and coaching, Parker notes.
This progress has contributed to large features in power effectivity over the previous eight years, which is a part of the explanation AI is now in a position to deal with more and more complicated issues.
What Is Accelerated Computing?
At its core, accelerated computing is about doing extra with much less.
It entails utilizing specialised {hardware} — like GPUs — to carry out duties sooner and with much less power.
This isn’t simply theoretical. “The change in effectivity is actually, actually dramatic,” Parker emphasised.
“In case you examine the power effectivity for AI inference from eight years in the past till at the moment, [it’s] 45,000 instances extra power environment friendly,” Parker mentioned.
This issues as a result of as AI turns into extra widespread, the demand for computing energy grows.
Accelerated computing helps firms scale their AI operations with out consuming large quantities of power.
This power effectivity is essential to AI’s potential to deal with a few of at the moment’s greatest sustainability challenges.
AI in Motion: Tackling Local weather Change
AI isn’t simply saving power — it’s serving to to combat local weather change.
“AI and accelerated computing normally are game-changers with regards to climate and local weather modeling and simulation,” Parker mentioned.
As an illustration, AI-enhanced climate forecasting is changing into extra correct, permitting industries and governments to arrange for climate-related occasions like hurricanes or floods, Parker defined.
The higher we will predict these occasions, the higher we will put together for them, which suggests fewer assets wasted and fewer injury performed.
One other key space is the rise of digital twins — digital fashions of bodily environments.
These AI-powered simulations enable firms to optimize power consumption in actual time, with out having to make pricey modifications within the bodily world. In a single case, utilizing a digital twin helped an organization obtain a ten% discount in power use, Parker mentioned.
Which will sound small, however scale it throughout industries, and the affect is large.
AI can be taking part in a vital function in growing new supplies for renewable power applied sciences like photo voltaic panels and electrical automobiles, serving to speed up the transition to wash power.
Can AI Make Knowledge Facilities Extra Sustainable?
AI wants knowledge facilities to function, and, as AI grows, so does the demand for computing energy. However knowledge facilities don’t must be power hogs. Actually, they are often a part of the sustainability resolution.
One main innovation is direct-to-chip liquid cooling. This know-how permits knowledge facilities to chill their methods far more effectively than conventional air-con strategies, which are sometimes energy-intensive.
“Our advisable design for the info facilities for our new B200 chip is concentrated all on direct-to-chip liquid cooling,” Parker defined. By cooling immediately on the chip stage, this technique saves power, serving to knowledge facilities keep cool with out guzzling energy.
As AI scales up, the way forward for knowledge facilities will depend upon designing for power effectivity from the bottom up.
Which means integrating renewable power, utilizing power storage options and persevering with to innovate with cooling applied sciences.
The aim is to create inexperienced knowledge facilities that may meet the world’s rising demand for compute energy with out growing their carbon footprint.
“The compute density is so excessive that it makes extra sense to spend money on the cooling since you’re getting a lot extra compute for that very same single direct-to-chip cooling factor,” Parker mentioned.
The Position of AI in Constructing a Sustainable Future
AI isn’t just a software for optimizing methods — it’s a driver of sustainable innovation.
From enhancing the effectivity of power grids to enhancing provide chain logistics, AI is main the cost in decreasing waste and emissions.
“AI, I firmly consider, goes to be the perfect software that we’ve ever seen to assist us obtain extra sustainability and extra sustainable outcomes,” Parker mentioned.
Register for the NVIDIA AI Summit DC to discover how AI and accelerated computing are shaping the way forward for power effectivity and local weather options.