Smarter mobile chipsets key to democratization of Artificial Intelligence
Source – livemint.com
New Delhi: Artificial Intelligence (AI) may be a hot topic in tech circles, but its real benefits will be served when it has been truly democratized, making it available for each individuals. While we are not so distant from a personal AI butler, tech firms are already making efforts in that direction. The first and critical aspect of the rollout is deepening the AI integration to the core of mobile phones.
From Samsung Electronics Co., LG Electronics Inc., Huawei Technologies Co. Ltd to Xiaomi Inc., a number of old and new smartphone players are leveraging Artificial Intelligence in their newer phones in one way or another. And then, we have companies like Google (Alphabet Inc.) and Microsoft Corp., which are incorporating AI in their software to deliver smarter products like Google Photos and Windows OS. Besides, we now have Samsung’s Bixby, Microsoft’s Cortana to Amazon.com Inc.’s Alexa—virtual assistants that are a core part of modern mobile computing devices.
The most critical aspect of implementing Artificial Intelligence in your mobile devices is baking the technology into the chipsets.
Before launching widely acclaimed AI phones like Mate 10, Huawei introduced Kirin 970 processor in September last year. Touted as the world’s first mobile AI computing platform, Kirin 970 can use on-device AI to enhance users’ vision and hearing. “The core of the AI technology is the ability to process mass data. Traditional computing architectures centered on the CPU, GPU, and DSP are no longer sufficient to meet overwhelming demands for computing performance in the AI era,” Huawei explains on its website.
“On-device AI processing capabilities are especially important because smartphones need to process data in real time and at any time, while protecting users’ private data as effectively as possible. As a result, on-device performance is the greatest obstacle for further development of mobile AI technology.”
With more brands looking to integrate AI features into their phones, AI chipsets will become highly critical. While Huawei may have hogged the limelight for mobile AI chips, Qualcomm Technologies Inc. has been making giant strides in this direction as well.
Qualcomm is eyeing the next mobile computing frontier with advancements in hardware architectures like vector processing on Snapdragon Hexagon DSP. Qualcomm has been using on-device AI in last three-generations of its Snapdragon mobile chipsets.
Earlier this year, Qualcomm introduced the Qualcomm AI Engine, which comprises several hardware and software components to accelerate on-device, AI-enabled user experiences on select Qualcomm Snapdragon mobile platforms. The AI Engine is supported on Snapdragon 845, 835, 821, 820, 710 and 660 mobile platforms, as well as the recent QCS605 and XR1 platforms.
“Qualcomm is making AI chips more efficient through hardware architecture advancements on all Snapdragon cores and a large investment in software optimizations,” said Gary Brotman, director of product management, Qualcomm Technologies.
“In fact, through software alone, we have been able to improve AI performance across the Snapdragon SoC portfolio by more than 200% over the past year,” he said.
Apart from energy efficiency and user experience, mobile AI chips are also betting big on the security and privacy, a growing concern for modern computing devices.
“Qualcomm Technologies has a rich history of hardware based security on our mobile processors. We believe it is fundamental to our security strategy. We will continue to work with our partners as they begin to implement software security solutions that utilize AI,” said Gary.