A new device shows how man and machine can come together to become unstoppable.
Technology is advancing at such a rapid speed that the world and its energy infrastructure are struggling to keep up.
Artificial intelligence (AI) is the central driving force behind these swift advancements, and its demands seem to be endless.
Will this innovative piece of engineering finally help us keep pace with modern technology and its ravenous hunger?
How the insatiability of AI is depleting resources
There is nothing quite as satisfying as seeing something grow into what it was always meant to be.
This feeling is, of course, applicable regarding modern-day technology.
Some of the generations alive today grew up without most of the innovations that are common today.
This includes the internet, social media, smartphones, and television.
Now, these concepts are a given, and central to them all these days is AI.
The transition happened almost effortlessly, but no one could have predicted how heavy the consequences would be.
This growth has now reached a point of no return. Globally, grids are struggling to feed AI’s “ravenous hunger.”
One commercial AI model’s training uses as much power as hundreds of households consume annually. The powerhouse behind AI, namely, data centers, is also extremely energy-intensive.
Naturally, these hubs and their hardware perform intense processing tasks. To prevent them from overheating, cooling systems require millions of gallons of water.
What will address this hunger and thirst?
Renewable energy sources alone cannot fill the gap
We are in a digital age, and have become increasingly important to protect critical infrastructure and operations.
To keep these systems from failing due to insufficient power, nations are investing in and expanding renewable energy capacity.
On paper, it seems like a concise answer, as it also addresses climate goals by keeping carbon footprints low.
Combining wind and solar energy in hybrid parks seems ingenious, as they complement each other to ensure uninterrupted clean power.
However, even hybrid systems fall short. In cases with no wind at night, dependency falls on old-school infrastructure not designed to meet AI demands.
Upgrading existing systems takes time that we do not have.
Fortunately, researchers from the University of Cambridge dedicated their time to finding a solution.
In their study, instead of offering sustainable energy solutions for AI, they decided to make AI itself more sustainable.
Stilling the hunger by adding some plasticity
You can find the study “HfO2-based memristive synapses with asymmetrically extended p-n heterointerfaces for highly energy-efficient neuromorphic hardware” published in Science.
AI is not going anywhere soon, and its role in the oil and gas industry is growing. But there is still a fundamental missing piece to the puzzle.
When a standard computer moves data between its memory and processor, it loses a lot of energy. This is called the “von Neumann bottleneck.”
So, the Cambridge scientists turned to a more human side of things to solve the energy puzzle.
Copying the human brain for seamless, efficient processing
A human brain stores and processes information in one location thanks to synaptic plasticity.
To help technology do this, the scientists engineered an “interfacial memristor.” It is made from specialized hafnium oxide, barium, and strontium.
The device thus functions like a “synthetic synapse.”
The “in-memory computing” prevents overheating, and ultra-low power operations cut AI energy consumption by 70%.
By making high-tech machines more “human,” we effectively fill the gap the world has been struggling with.
To step forward into a greener future, the shift should not just focus on technology producing clean power. Instead, it should also focus on how we use it.
Perhaps this human brain-inspired device will finally help data centers and the energy sector embrace true sustainability.
