Qualcomm Showcases Machine Learning On A Snapdragon SoC

October 1, 2015 - Written By Muni Perez

Machine learning is advancing in a fast pace and the technology is becoming very accessible for common users. An example of that is Google Now, the personal digital assistant, that learns things about a user and is capable of taking decisions on what information you are most likely to find useful. However, in order for a computer system to make real-time calculations of general and (for us) trivial daily things, a large computational structure is needed, including petaflops of processing power, enormous amounts of RAM and really large amounts of electricity. As such, with the technology existent today, you simply cannot run something like a Google Now query on your smartphone, it doesn’t matter how powerful it is. You may have noticed that in order to use the assistant, you need internet. This is because the information provided is sent to Google’s servers, processed and then a final result is sent back to you. Qualcomm is the world’s leading manufacturer of mobile processors and their research division has been working on bringing this kind of computing power to mobile devices.

Earlier this year they have showcased the Zeroth Cognitive Computing Platform, a machine learning system that runs on regular Snapdragon System-on-a-Chip (SoC), the one you are probably using to read this article and browse AH’s great content. The objective is to create an independent cognitive network that doesn’t need to send data to an external server. One application showcased by the company is a visual perception algorithm that has a wide range of uses. Imagine your device being able to identify someone on your photos, or a security system that can differ your pet from an intruder (without the use of internet), or a car safety system that can alert the driver for a distracted pedestrian or a dangerous situation on the road. The company released a video where the technology is being used in a drone, allowing it to scan the environment ahead in order to determine if it is safe to fly. It can detect trees and structures and alert the pilot about the course ahead.

We are still a few years away from having such technology on our hands, but the future is really promising. Not needing an internet has two advantages – the first and obvious one is that a user will still be able to use great artificially intelligent apps even without a cellular or Wi-Fi connection and it can save overall bandwidth for more important applications like IoT networks, which are set to grow exponentially in the next few years. Connected cars, smart sensors in a city and every other smart home devices will need internet to communicate, and having a network-independent cognitive system working on small mobile processors will certainly benefit the overall infrastructure. Check out the video below and see for yourself what Qualcomm is preparing for us.