While much attention these days is on the cloud, Qualcomm is focused on putting more intelligence into the device. In fact, Qualcomm has made artificial intelligence (AI) a priority, second only behind 5G, according to Gary Brotman, director of product management, artificial intelligence/machine learning at Qualcomm Technologies.
Brotman’s team is focused on neural networks on devices, where AI is a term that’s designed to capture anything to do with helping a device mimic the processes and patterns of the human brain. That sounds like a lofty job, but they’re making striking inroads.
Brotman presents a good argument for why you wouldn’t want everything in the cloud. With a lot of recognition-type things moving to the device—think facial recognition—that involves a lot of personal data that you don’t want in the ether. Keeping it on the device can avoid the round trip to the cloud and help protect privacy.
The ability to do this kind of processing on the phone is thanks in no small part to Snapdragon. “Snapdragon’s been designed to do this stuff for three generations,” Brotman said. “We’ve really been optimizing for neural network execution on mobile and other edge devices since Snapdragon 820.”
Just before Mobile World Congress 2018 kicked off, Qualcomm introduced its Qualcomm AI Engine, which is comprised of several hardware and software components to accelerate on-device AI-enabled user experiences. The AI Engine will be supported on Snapdragon 845, 835, 820 and 660 mobile platforms, with advanced on-device AI processing found in the Snapdragon 845.
AI is also playing a role in making voice assistance more lifelike. Assistants like Cortana, Siri and Alexa are popular now, but they’re not the same as talking with a human. Natural language processing enables more free-flowing conversation instead of key word requests. “That’s starting to happen now” among developers that are using the Qualcomm solution, Brotman said.
“When you can bring the conversation down or the dialog down to the device level as opposed to joining up to capture what you need from the cloud, the gaps and the awkwardness disappear” and you’re having a conversation rather than a command-and-control kind of dialog, he said.
It’s not just about recognizing a voice but a particular voice. What if a voice-activated persona could not only sound more lifelike but actually be able to “read” a person’s emotions? For example, if you come home after a long day at work, the “assistant” can figure out that you need some calm and relaxing music as soon as you enter the room. It’s those kinds of things that Qualcomm’s AI unit aspires to enable.
Qualcomm’s designs in AI are not as well-known as its work in 5G, but Brotman said AI/machine learning is impacting everything, from product development to back offices. Snapdragon AI cloud collaborations include Tencent, Baidu, Google, Facebook, Microsoft and Amazon. The Snapdragon AI developer ecosystem includes SenseTime, Thundercomm, Elliptic Labs, Face++ and Uncanny Vision.
Smartphone manufacturers have also taken advantage of AI Engine components on Snapdragon mobile platforms to accelerate AI applications on their devices, including Xiaomi, OnePlus, vivo, Oppo, Motorola, Asus, ZTE, Nubia, Smartisan and Blackshark, some of which are planning to optimize AI applications using the AI Engine on future flagship smartphones.
“There’s a role for the device. It has enough compute to be able do quite a bit, to make the consumer experience seamless and private, and there’s still the cloud,” Brotman said, noting they’re not trying to take the cloud out of the equation, but “our belief is devices should be autonomous and intelligent in their own right.”
Given that 8.5 billion smartphone shipments are forecast between 2017 and 2021, that makes mobile a pervasive platform for AI and a pretty solid bet for Qualcomm.