Slide 1: 5G+AI: The ingredients fueling tomorrow's technology innovations
February 6, 2020 @qualcomm_tech 5G+AI: The ingredients fueling tomorrow’s technology innovations
Slide 2: Delivering on the 5G vision
Delivering on the 5G vision $13.2 trillion in goods & services by 2035* 2
Slide 3: The intelligent wireless edge
New experiences Privacy/security New verticals Immediacy Processing over 5G The intelligent Efficiency Edge cloud wireless edge On-device Customized/ local value Reliability On-device AI processing augmented by edge cloud Private/public networks Personalization 3
Slide 4: Process data at the source to scale AI and make sense of a digitized world
Process data at the source to scale AI and make sense of a digitized world Past Today Cloud-centric AI Partially-distributed AI AI training and AI inference Power-efficient in the central cloud on-device AI inference On-device Future Fully-distributed AI With lifelong on-device learning 4
Slide 5: Applying AI to overcome wireless challenges
Applying AI to overcome wireless challenges + 5
Slide 6: Applying AI to solve difficult wireless challenges
Applying AI to solve difficult wireless challenges Deep wireless domain knowledge is required to optimally use AI capabilities Wireless challenges AI strengths Determining appropriate Hard-to-model problems representations for hard-to- model problems Computational infeasibility of optimal solution F
Slide 7: AI enables intelligent 5G network management
AI enables intelligent 5G network management Enhanced service quality Simplified deployment Better mobility management, user More capable Self Organizing localization, and user behavior and Networks (SON) for e.g., mmWave demand prediction network densification Higher network efficiency Improved net
Slide 8: On-device AI improves the 5G end-to-end system
On-device AI improves the 5G end-to-end system Enhanced device experience More intelligent beamforming & power management improve throughput, robustness, and battery life On-device Improved system performance AI On-device inference reduces network data traffic for more efficient mobility and spectru
Slide 9: Radio awareness
Spectrum sensing and access Predict activities of other devices for more efficient access and better scheduling to improve 5G system performance Contextual awareness Radio awareness Use device context (e.g., position, velocity, or in- car) derived from RF, sensors, traffic activities to Achieved by
Slide 10: On-device AI enhances 5G device experience
On-device AI enhances 5G device experience Agent Better beam management Incorporate location, velocity, other aspects State of environmental and application awareness M strongest to improve robustness and throughput Action RSRPs Select Tx/Rx On-device beams Reward AI Throughput, More power saving po
Slide 11: On-device AI improves 5G system performance
On-device AI improves 5G system performance Better link adaptation Position-aware interference prediction can improve overall system throughput and spectral efficiency AI inference result More seamless mobility Reduced network loading Device-centric mobility utilizes on-device On-device AI inference
Slide 12: Applying AI for enhanced 5G air interface efficiency
Applying AI for enhanced 5G air interface efficiency Example: for uplink transmissions Implementing a neural network framework for CSI1 on non-linear temporal encoding & decoding Improving system Downlink Data or Data or spectral efficiency CSI CSI Reconstructed channel control control downlink chan
Slide 13: Applying AI for contextual awareness & environmental sensing
Applying AI for contextual awareness & environmental sensing More accurate device positioning Motion and gesture detection RF sensing Multi-path detection NLOS Blockage LOS Learning device position over time Sensing changes in environment to without prior knowledge with RF infer location and type of
Slide 14: AI research areas to enhance 5G
Full duplex Spectrum sensing Channel estimation Radio resource allocation Power saving Security MIMO detection Environmental sensing AI research areas to Device non-linearity enhance 5G Contextual awareness Positioning Vehicular communications TCP optimization Beam management and optimization 14
Slide 15: Applying 5G to enhance AI-powered experiences
Applying 5G to enhance AI-powered experiences + 15
Slide 16: 5G + AI deliver enhanced services and experiences
5G + AI deliver enhanced Distributed processing, like boundless XR services and experiences Distribute AI processing between the device and Real time assisted cloud over wireless services like voice UI Distributed functionality On-premise control for ultra-low latency Central cloud Edge cloud Device
Slide 17: Personalized shopping through boundless AR
Personalized shopping through boundless AR AI processing split between device and edge cloud 17
Slide 18: The edge cloud augments on-device processing for boundless XR
The edge cloud augments on-device processing for boundless XR XR headset Cloud Edge cloud Rendering, perception, prediction, Less latency sensitive content Rendering, scene understanding, personalization (e.g. recorded streaming video) classification, recommendation, … Head pose, image, … On-device
Slide 19: Intuitive virtual assistants through vastly improved voice UI
Intuitive virtual v assistants through vastly improved voice UI Designed to be: Always-on Conversational Personal Private 19
Slide 20: Distributed computing enables a responsive voice UI
Distributed computing enables a responsive voice UI 5G low latency allows AI tasks to be split between the device and cloud Less privacy and Continuously personalization learns based on personal Automatic speech Voice information and recognition AI agent activation acts intuitively with immediacy Se
Slide 21: The reconfigurable factory of the future will thrive in the 5G+AI era
The reconfigurable factory of the future will thrive in the 5G+AI era 21
Slide 22: Enhanced mobile broadband, Massive IoT, Ultra-reliable low-latency
Enhanced Massive mobile broadband IoT Computer vision Sensors Head mounted display Handheld terminal Ultra-reliable low-latency Wireless edge analytics Automated guided vehicle (AGV) Industrial robot 22
Slide 23: Distributed learning over wireless
Distributed learning over wireless 23
Slide 24: Distributed learning for an intelligent world
Distributed learning for an intelligent world 24
Slide 25: Devices generate and possess massive amounts of data
Devices generate and possess massive amounts of data Off-device data On-device data Sensor data Cloud data/rich content Calendar Environment Fingerprint Ambient light Compass IoT data Messaging Camera Microphone Temperature Humidity Sensor fusion Apps Gyroscope Pulse Eye tracking C-V2X and processin
Slide 26: Distributed learning over 5G is the way to scale intelligence
Distributed learning over 5G is the way to scale intelligence 1 A state-of-art global AI 2 Each device collects model is distributed to personal data and performs devices on-device training D1 D2 ... Cloud DN Scale Personalization Privacy Processing is spread over Model customized based on your Raw
Slide 27: Distributed learning over 5G is the way to scale intelligence
Distributed learning over 5G is the way to scale intelligence 4 The cloud updates the AI 3 Each device: model and then sends it to 1. Adds perturbation to the the devices again parameters of the AI model D1 2. Compresses the parameters 3. Encrypts them D2 4. Sends the update to the cloud ... Cloud 5
Slide 28: Research directions for distributed learning over wireless
Optimizing the communication network Compressing information sent on the uplink and downlink Tackling statistical heterogeneity in data Smartly combining model updates from a broad distribution Research directions Privacy for distributed learning Model parameter perturbations with privacy guarantees
Slide 29: Uniquely positioned to power the intelligently connected future
X55 Qualcomm Ventures AI Fund Foundational Qualcomm AI 5G + AI Advanced Ecosystem R&D Research technology silicon investment leadership Uniquely positioned to power the intelligently connected future Qualcomm AI Research is an initiative of Qualcomm Technologies, Inc. Qualcomm Snapdragon is a produc
Slide 30: AI is making 5G better — in the network and on the device
AI is making 5G better — in the network and on the device On-device AI processing is essential, and even more so in the 5G era 5G is empowering a new computing paradigm: distributed learning over wireless 30
Slide 31: Questions?
www.qualcomm.com/ai www.qualcomm.com/5g www.qualcomm.com/news/onq BLOG @qualcomm_tech Questions? YouTube Connect with Us SlideShare 31
Slide 32: Thank you
Thank you Follow us on: For more information, visit us at: www.qualcomm.com & www.qualcomm.com/blog Nothing in these materials is an offer to sell any of the References in this presentation to “Qualcomm” may mean components or devices referenced herein. Qualcomm Incorporated, Qualcomm Technologies,