Here Comes the Sun: A New Wave of Solar-Powered AI at the Edge
Mitchell Gracie posted on February 19, 2019 |
The new ASIC technology unveiled by Xnor is solar-powered, the size of a quarter, and is light enough to be lifted by a helium balloon. (Image courtesy of Xnor.)
The new ASIC technology unveiled by Xnor is solar-powered, the size of a quarter, and is light enough to be lifted by a helium balloon. (Image courtesy of Xnor.)

Cheaper, better, faster, smaller. For decades, those four words—not to be confused with a hit Daft Punk song—have both driven fear into developers and driven sales. However, as the energy burdens for the Internet of Things (IoT), cloud computing, crypto currencies and artificial intelligence (AI) increase, a fifth word is necessary: greener.

Xnor.ai (Xnor) isn’t scared of the “greener” challenges facing industries today, and the unveiling of its new application-specific integrated circuit (ASIC) technologies proves so.

“Power will become the biggest bottleneck to scaling AI,” said Ali Farhadi, co-founder of Xnor. “What Xnor has proved today is that it is now possible to run AI inference at such low power that you don’t even need a battery. This will change not only the way products are built in the future, but how entire cities and countries deploy AI solutions at scale.”

Battery-less and the size of a quarter, ASIC technology leverages solarcell technologies to deploy state-of-the-art deep learning models in the field without the need for voluminous batteries or power-hungry GPUs.

“There’s a reason we don’t see AI in our everyday life. Why is that? To me, the main reason is cloud dependency,” continued Farhadi. “Just think about how much power we are consuming on running traditional AI algorithms in the cloud.”

By localizing power burdens away from the cloud and toward the fog—that is, a shift from central processing to edge processing—Xnor expects to see faster application of AI.

Consumer wearables, autonomous vehicles and agriculture are some of the many sectors that Xnor sees primed for their clever hardware architecture and new machine learning algorithms.

“For every device that is processing an algorithm, that [action] is drawing energy from a data center. When you think about the number of devices and the number of data centers, the impact on our carbon footprint is enormous,” elaborated Sophie Lebrecht, senior vice president of strategy and operations at Xnor.

The reallocation of power spent on processing makes sense for businesses. A Natural Resources Defense Council (NRDC) study from 2015 informs us why this is the case. With American data centers expected to consume over 130 billion kWh of electricity each year by 2020, there exists an opportunity cost associated with remotely processing AI and machine learning algorithms. Not only does the abuse of the cloud “[cost] American businesses $13 billion annually in electricity bills”, but—according to the NRDC—it “[emits] nearly 100 million metric tons of carbon pollution per year.”

The Technology

Light enough to float while tethered to a helium-filled balloon, the new ASIC technology is embedded with a camera that allows it to intelligently detect visual objects—including humans—all while being solely powered by a simple solar cell. Without the need to charge the device, scaling AI technologies gets faster and better, according to Xnor.

Xnor is known for combining progress from machine learning, computer architecture and code optimization to innovate better AI solutions that can then be integrated onto low-cost Raspberry Pi computers. (Image courtesy of Xnor.)
Xnor is known for combining progress from machine learning, computer architecture and code optimization to innovate better AI solutions that can then be integrated onto low-cost Raspberry Pi computers. (Image courtesy of Xnor.)

Using energy in the magnitude of microjoules per inference and state-of-the-art IoT connectivity protocols, the current version can harvest and analyze video content by identifying, classifying and codifying objects within the view of its camera and upload processed data wherever it is needed, in real-time.

“With technology this low power, a device running on only a coin cell battery could be always on, detecting things every second, running for 32 years!” claimed Saman Naderiparizi, head of hardware engineering at Xnor.

Moreover, security isn’t sacrificed. Xnor claims that photo and video data are never distributed, only the analyzed frames, guaranteeing privacy.

It is an exciting time for green technologies, sensors and machine learning, and, with this announcement, Xnor hopes to see all converge, finding cheaper, better, faster, smaller and greener solutions for everyone.


Recommended For You