Physicist uncovers critical flaw in modern GPUs, presents revolutionary new alternative: 'People are most interested in the energy-efficiency angle'
Imagine a future where the devices we rely on every day — from smartphones to smart home appliances — are powered by technology that uses a fraction of the energy consumed by today's computer chips.
A new circuit designed by researchers at the University of Pennsylvania could turn that vision into reality, according to MIT Technology Review.
So, what is this game-changing innovation?
It's a simple assembly of 32 resistors that can be trained to sort data, much like the machine learning models that underpin today's artificial intelligence. But unlike the power-hungry GPUs those models run on, this humble circuit learns in a radically different — and much more efficient — way.
The key lies in how the circuit encodes information. Rather than rigid 1s and 0s, it uses variable voltages — an approach known as analog computing. Analog fell out of favor decades ago, but researchers have long known it offers superior energy efficiency to digital at scale.
Aatmesh Shrivastava, an electrical engineer at Northeastern University, affirms this advantage of analog, stating: "The power efficiency benefits are not up for debate," per MIT Technology Review.
Watch now: Uber-wealthy couple makes unprecedented move with $300 million land
The Penn team estimates their design could be 10 times more efficient than cutting-edge AI chips.
Mimicking the human brain, the resistors in the circuit adjust their values based on external feedback, without any top-down instructions. This neuromorphic approach eliminates the need for separate processor and memory components, which suck up significant power in conventional computers.
In trials, the circuit has already proved it can classify data with 95% accuracy and perform fundamental machine-learning operations. With further development to refine the design and scale it up, this unassuming prototype could mark a turning point in the race to create faster, greener computing.
According to lead researcher Sam Dillavou, the technology has piqued the interest of several companies. "People are most interested in the energy-efficiency angle," he noted, per MIT Technology Review.
While a commercial neuromorphic device based on the Penn team's work may still be years away, their outside-the-box thinking shines a light on the path to a more sustainable computing future — one that could make our daily digital demands a little easier on the planet.
Join our free newsletter for weekly updates on the coolest innovations improving our lives and saving our planet.
Physicist uncovers critical flaw in modern GPUs, presents revolutionary new alternative: 'People are most interested in the energy-efficiency angle' first appeared on The Cool Down.
Welcome to Billionaire Club Co LLC, your gateway to a brand-new social media experience! Sign up today and dive into over 10,000 fresh daily articles and videos curated just for your enjoyment. Enjoy the ad free experience, unlimited content interactions, and get that coveted blue check verification—all for just $1 a month!
Account Frozen
Your account is frozen. You can still view content but cannot interact with it.
Please go to your settings to update your account status.
Open Profile Settings