Neural networks, the invention of chips that can learn and think on their own, began back in the early days of AI research. Now referred to as “deep learning systems”, these neural networks are how Facebook is able to recognize your mom’s face in a photo you post. It is also how automobile computers can recognize street signs and signals (self-driving cars), how robots or drones can make quick commands and anticipate actions, or how a device can know it’s your voice to log into a private account. The intention for these brain-like chips is to get them small enough to to feasibly work in our mobile devices. The burden of how large these graphics processors have to be, and their large energy demand, has been the reason for the wait. However, the researchers at MIT want to go ahead and clear up this issue by offering a chip, dubbed Eyeriss, that “promises neural networks in very low-power devices”.
Despite the fact that large GPUs work using multitudes of cores, Eyeriss’ 168 cores don’t necessarily mean the consumption of large amounts of energy. In fact, MIT researchers claim it consumes a whopping 10 times less power and takes a fraction of the resources than the same little GPU you would find in your smartphone. Eyeriss will process and perform all of its brain-like functionality without needing to tap into other data sources, such as Wi-Fi or the cloud. In the instance of a self-driving vehicle, this method of on-board brain power would be helpful if the driver were to, say, find himself in a remote area with little to no internet (PC World).
Internet of Things (IoT) devices will also benefit from Eyeriss. By using on-board AI algorithms, these IoT devices could “make important decisions locally, entrusting only their conclusions, rather than raw personal data, to the internet”; not to mention the use that could be given to battery powered autonomous robots (Cellular News).
The new chip has already been demonstrated to the public during the ISSCC (International Solid-State Circuits Conference), where MIT researchers showed off image recognition skills. We don’t know when to expect the Eyeriss technology implemented into our favorite smartphone or smartwatch. However, when it does happen, it is going to be a huge benefit for those of us who would love for our device to have that extra brain power.
Read Related Articles
- AMD’s FreeSync 2 Completely Ups the Ante With Lower Latency, and the Addition of HDR
- Razer’s Project Ariana Takes Gaming to Immersive New Levels
- The Linksys Velop Wants to Beat Competition With a More Powerful Wi-Fi System
- The Best Windows Laptop of 2016 Just Went Convertible
- The New Military Grade Acer Chromebook for Schools Can Withstand a Beating, or a Spill!
- Announced at CES: Great Lineup of Asus Devices Include Intel’s Latest Kaby Lake CPUs