tech2 News StaffMar 21, 2017 14:05:30 IST
Artificial Intelligence truly is the stepping-stone into the future. It’s hard to find any ‘smart’ device these days without some sort of digital assistant thrown in. Any new app or feature usually has some tie-in with AI.
Digital assistants like Siri, Alexa and Google Assistant rely on the cloud to function correctly. Any real time application of AI usually requires some sort of dedicated hardware.
Arm today announced the release of its new DynamIQ platform, technology that will soon find its way into the next generation of mobile processors. The platform integrates the functionality required for an AI / machine learning implementation into Arm’s traditionally power efficient chip design.
But why is this a big deal?
Arm processor cores are what you find in just about every low-powered IoT device and every smartphone in the market today. Companies like Qualcomm license the architecture from Arm and then design an SoC on around it.
Unlike desktop CPUs, Arm chips use what is called the RISC (Reduced Instruction Set Computing) model. Processors process instructions using transistors. RISC instructions are short, simple and easy to execute. Because they’re simple, they require fewer transistors to process.
The simplicity of the instructions means that the processor can be smaller and more power efficient, and that’s the key to Arm’s success. Arm chips are extremely power efficient and thus, ideal for use in low power mobile applications like IoT and mobile devices.
The problem with RISC is that it’s not very flexible. To oversimplify things, say you want to perform a calculation of 9x134. Hypothetically, assume that RISC doesn’t allow for the multiply function. The processor would need to perform an addition instruction 134 times. Every single time, the sum of the previous figures would need to be stored and returned from memory and the number of addition instructions will need to be rechecked. Each instruction is faster than a single multiply instruction, but together, the process is much longer.
The bottom line is that Arm chips weren’t originally designed for processing the instructions that pertain to machine learning. DynamIQ fixes this by incorporating dedicated instructions for AI.
An efficient architecture for processing AI logic means that AI can now be processed on your device. This also means that your digital assistant, for example, will not need to access the cloud to process even simple voice commands.
Arm is positioning DynamIQ for smart cars as well, and this makes sense. Smart vehicles will need to be self-sufficient and take decisions in real-time; you can’t afford to wait 2-3 seconds as your car decides whether it should apply the brakes or not.
The company also points out that AI processing on your smart device means more privacy as the data need never leave your device.
It must be noted, however, that Qualcomm already integrates a machine learning module into some of their SoCs and have designed a machine learning platform called Zeroth.
We've reached out to Qualcomm for comment and will update the story with the same.
Find latest and upcoming tech gadgets online on Tech2 Gadgets. Get technology news, gadgets reviews & ratings. Popular gadgets including laptop, tablet and mobile specifications, features, prices, comparison.