OnQ Blog

AI Day 2019: Removing barriers and accelerating progress with AI [video]

“We are moving toward a future where everything will be connected and intelligent,” Cristiano Amon, the President of Qualcomm, said as he kicked off this year’s AI Day in San Francisco.

At Qualcomm Technologies, we’re well-positioned to power this future. We’ve invented many of the breakthrough technologies that fueled the last 30 years of innovation and have collaborated with cloud companies and software developers around the world. We’ve also heavily invested in AI R&D for over a decade. Our technologies help enable the current Invention Age, a time of rapid acceleration in mobile, IoT, and automotive, networking, cloud infrastructure, and beyond. By combining our AI know-how with the benefits of 5G, we’re introducing new opportunities for innovation.

Here’s an overview of what we announced today — the technologies that will help make AI ubiquitous in our everyday experiences with mobile and beyond:

Cloud AI

When it comes to the cloud, there’s an increasing demand for AI inference to address peak performance coupled with power efficiency. The era of big heat sinks attached to CPUs to do this part of cloud AI will shift toward a new breed of AI accelerators. To meet this demand, we applied our expertise in designing for performance with power efficiency, scale, process nodes, and signal processing to engineer the Qualcomm Cloud AI 100.

The Cloud AI 100 is designed to offer a more than 10X performance improvement over competitive AI inference solutions that span CPUs, GPUs, and FPGAs. It’s built from the ground up, using a 7nm process node designed to support leading performance and power advantages.

With the Cloud AI 100, developers will be supported with a full stack of tools and frameworks, so they can create trained neural net models that can be deployed across the edge cloud and edge devices. It creates support for cutting-edge software stacks, including TensorFlow, Keras, PyTorch, Glow, and ONNX. The edge cloud augments on-device processing to enable low-latency responses, low-power processing, low bandwidth costs, and small storage space.

By expanding the distributed AI processing workloads, we’re enhancing user experiences, such as computer vision applications, personal assistants for natural language processing and translations, Extended Reality (XR), and personalized recommendations and content.

AI and our mobile platforms

Today, the Qualcomm Snapdragon 730, Snapdragon 730G, and Snapdragon 665 Mobile Platforms debut, bringing in-demand, next-generation features to new tiers of devices. Among them is on-device AI delivered via our third (for the Snapdragon 665) and fourth generation (for the Snapdragon 730/730G) Qualcomm AI Engines.

Our new Snapdragon 730 Mobile Platform truly takes mobile experiences up a notch. It features the fourth generation Qualcomm AI Engine, which optimizes user experiences with a heterogeneous computing approach. The Qualcomm AI Engine includes the newly designed Qualcomm Hexagon Tensor Accelerator for dedicated AI processing for the camera, gaming, voice, and enhanced mobile payment security. End users will be able to send secure mobile payments quickly, capture better low-light photography, interact with smart voice recognition technology, and so much more.

The Snapdragon 730 is also the first in the 7 Series to feature a Computer Vision ISP (CV-ISP), which delivers 4X faster computer vision processing compared to its predecessor. When paired with the camera, it backs advanced settings, such as 4K HDR, super resolution, super night shot, and object detection. With these options at their disposal, users can not only take awe-inspiring photos, but also über-impressive videos, allowing them to express themselves in unique ways.

For those who desire uncompromising gaming performance, the Snapdragon 730G is equipped with select Snapdragon Elite Gaming features. In terms of graphics, the platform’s Qualcomm Adreno GPU renders 25 percent faster than the Snapdragon 710 and 15 percent faster than the Snapdragon 730. It also taps its Game Jank Reducer to help eliminate stops and stutters that may occur on the device’s screen. And to fully immerse the user, True HDR Game Rendering displays the game in over a billion shades of color, presenting a gaming environment that’s almost true-to-life. When these Snapdragon Elite Gaming features work together, users can have a competitive edge against their opponents and enjoy the mobile game the way the developers intended.

In our Snapdragon 665 Mobile Platform, AI and the triple camera it enhances join forces to shoot high-quality photos. The platform is equipped with our third generation Qualcomm AI Engine, which processes AI up to 2X faster than the previous platform. It can capture 48MP snapshots and offers portrait mode, low-light night mode, and auto scene detect options.

AI and IoT

AI is fueling exciting innovation for the Internet of Things (IoT), and it’s driving a push toward high performance edge processing. Adding AI to the edge helps IoT devices lower latency, increase security, and lower bandwidth usage, allowing them to run smoothly and efficiently. A few AI-powered IoT use cases that benefit from higher edge processing are facial recognition, object and inventory tracking, scene classification, navigation and obstacle avoidance, and voice UI in today’s wearables, hearables, smart speakers, and home hubs and other types of objects in the future.

We’re helping customers and industries across the IoT to add vision and audio intelligence, connectivity (which includes 4G, 5G, Wi-Fi, and Bluetooth), low-power consumption, and high-performing edge processing to a variety of use cases. These benefits are enabling exciting new uses cases and inventions with applications in industry 4.0, smart displays, retail, smart speakers, home and enterprise robotics, smart cities, and enterprise and home security cameras.

Two examples of how we’re marrying AI and IoT are our latest Qualcomm Vision Intelligence 400 and 300 platforms, which are supported by our Qualcomm QCS605 and Qualcomm QCS603 SoCs. They’re purpose-built for IoT, using a 10nm FinFET process technology that integrates our advanced ISP, the Qualcomm AI Engine, and our heterogeneous compute architecture.

We simplified this heterogeneous computing by including easy-to-use SDKs in the Qualcomm Vision Intelligence Platform Series, so developers have the freedom to implement different AI use cases and can commercialize their innovations quickly. The platforms also support up to 2x2 802.11ac Wi-Fi with MU-MIMO, Bluetooth 5.1, Qualcomm 3D Audio Suite, Qualcomm Aqstic Audio Technologies, and Qualcomm aptX Audio.

With the Qualcomm Vision Intelligence 400 and 300 platforms, developers can harness Qualcomm Technologies’ superior connectivity and security technologies in their products.

AI and audio

Performance audio meets connectivity and AI in the recently announced Qualcomm QCS400 SoC. With its integrated Qualcomm AI Engine, far-field multi-keyword voice pickup with beamforming, and low-power voice detection, the Smart Audio SoC series supports faster response times for voice commands, more robust voice pickup in noisy environments, and local voice control than our previous generation technology. The QCS400’s on-device AI helps it to respond to a user’s voice even when the device isn’t connected to the internet — this means users can take their speakers with them into different rooms, in the backyard, or use them at the beach, without loss of the voice interface.

We engineered the QCS400 with robust, multi-channel connectivity that supports low-latency wireless audio, multiple speakers to be networked together around a home, and smart-home features like controlling lights or thermostat using Voice UI. And for developers, the QCS400 is designed to be compatible with third-party neural networking frameworks and the Qualcomm Neural Processing Engine SDK, as well as our ADK for other feature customization.

AI and automotive

AI is well on its way to transforming the automotive industry as well. Soon, cars may be able to recognize drivers by their faces through an AI-based camera to provide personalized experiences. Cars could also provide intuitive and natural interactions, understanding drivers’ voice and gestures to provide music or restaurant recommendations, just like a digital assistant would.

Cars are also expected to use AI-powered cameras for in-cabin monitoring and ultra-HD surround view monitoring. The cameras could recognize drivers’ facial expressions and detect drowsiness. They may also be designed to scan cars’ surroundings and alert drivers when the car in front of them abruptly stops.

Qualcomm Technologies has worked relentlessly to drive the development and commercialization of AI-enhanced technologies and 5G. Beyond today’s news announcements, we’re continuing to develop the edge cloud, so industries can take full advantage of the powerful combination of AI solutions running on a 5G network.

 

 

Qualcomm Cloud AI, Qualcomm Snapdragon, Qualcomm AI Engine, Qualcomm Snapdragon Elite Gaming, Qualcomm Adreno, Qualcomm Vision Intelligence Platform, Qualcomm Hexagon, Qualcomm 3D Audio Suite, Qualcomm Aqstic, Qualcomm aptX, Qualcomm QCS605, Qualcomm QCS604, and Qualcomm QCS400 are products of Qualcomm Technologies, Inc. and/or its subsidiaries.

 

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries ("Qualcomm"). Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries. The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.