AI WORKSHOP IIT DELHI 29 TH NOVEMBER 2019

AI WORKSHOP, IIT DELHI 29 TH NOVEMBER 2019 1 Satish Jamadagni

AI Hardware Three types of AI computing • Centralized environment, all calculations are done on one particular computer system, such as a dedicated server for processing data. • the single server model bottleneck • Distributed computing, not all transactions are processed in the same location, but that the distributed processors are still under the control of a single entity. • SETI • Decentralized computing, on its end, entails that no one single entity has control over the processing. • no leakage of sensitive data or attack on security

AI Hardware Nvidia dominant vendor of GPUs updating to support neural networks Googles Tensor Processing Unit (TPU) - 30 times faster , 80 times less power consumption Facebook and Intel Nervana Chip NNP‐I - optimized for trained algorithms AWS Inferentia - large workloads, lower latency, designed for inference, find patterns in large data Intel’s Myriad 2 AI Chip - Movidius enhanced low power AI for vision and imaging IBM’s 8‐Bit Analog chip - based on phase‐change memory; double the accuracy and consumes 33 x less energy Huawei’s Ascend 910 and Ascend 310 - AI Chips for datacentres, smartphones, smartwatches and Io. T

AI / ML / DL

AI Challenges Determinism Still the biggest challenge

Focus on 5 G – Network Operations (Big Data) Challenges

Focus on 5 G – Network Operations (Big Data) Opportunities

Focus on 5 G – Slice Analytics & Diagnostics Problem statement • Network slicing has additional dimensions wrt. RAN management automation • For resource allocation optimization and SLA assurance, slicing-aware prediction methods are required • (Question: Is closed-loop automation possible) Results Slice-aware Network Element (NE) state model: quantization of NE KPIs into a selected number of states States → Long-Short Term Memory (LSTM) Recurrent Neural Network (RNN) → State Prediction

Focus on 5 G – Radio Access Networks Source: Nokia

Focus on 5 G – Radio Access Networks

AI Standards – What's happening ETSI Network Intelligence Core Standard Group - ETSI ENI （ Experiential Networked Intelligence ）

AI Standards – What's happening ETSI ENI Po. C Example: project #1 - Intelligent Network Slice Lifecycle Management

AI Standards – What's happening ETSI ZSM (Zero-touch-network-service-management) architecture

AI Standards – What's happening NIST - US NIST has released a plan for prioritizing federal agency engagement in the development of standards for artificial intelligence (AI) per the February 2019 Executive Order on Maintaining American Leadership on Artificial Intelligence (EO 13859). The plan recommends the federal government “commit to deeper, consistent, long-term engagement in AI standards development activities to help the United States to speed the pace of reliable, robust, and trustworthy AI technology development. ”

AI Standards – What's happening India Source: https: //analyticsindiamag. com/ai-standards-us-china-india/

AI Standards – What's happening India- TSDSI Work Items on the following areas: RAN APIs for 5 G, The ORAN group is working on the RAN disaggregation models, in that context there will be a WI in TSDSI SG-N to provide a set of APIs for RAN control. IOT Data Analytics, APIs to facilitate data analytics in both homogeneous and heterogeneous IOT platform / service deployment scenarios (SGSS Work Item) Federated AI APIs (Schemas and semantics), The challenge of learning in a distributed cloud computing scenario needs to be addressed. This will be a Phase II activity for the on going cloud interworking activity (SGSS) Reliance Jio will champion all of the above activities

THANK YOU