From Cutting-Edge AI Research to Business Impact - What Everyone Ought To Know About Outcome Based Automation Confidential and Proprietary. © 2019 UST Global Inc 1 We Are The Right Size to be an Effective and Flexible Partner Bringing to you the benefits of our Fewer Clients More Attention model With 25,000 employees in 26 countries, we are big enough to scale, yet the right size to care Being privately-owned allows us to be flexible to your needs andConfidential go Beyond and Proprietary. the© 2019Contract UST Global toInc serve 2 Our global presence to ensure access to talent and seamless delivery 42 London 34 Leeds Operating Ireland Delivery Centers Geneva Centers Madrid Copenhagen Cologne Gdansk Netherlands Algeciras Barcelona Singapore Aliso Viejo Manila New York, NY Shanghai Norfolk, VA Penang Dayton, OH Tel Aviv Kuala Lumpur Leon Atlanta, GA Cyberjaya Guadalajara Bentonville, AR Taiwan Mexico City Toronto Hong Kong Costa Rica Brazil Trivandrum Kochi Coimbatore Bangalore Chennai Hyderabad Bhopal Pune Mumbai Sydney Major Delivery Centers Delhi Auckland Confidential and Proprietary. © 2019 UST Global Inc 3 USA India Poland Philippines Mexico Malaysia Spain Singapore China Taiwan Germany Australia Adnan Masood, Ph.D. - Chief Architect - Artificial Intelligence and Machine Learning at UST Global, Visiting Scholar at Stanford University, and Microsoft MVP (Most Valuable Professional) for AI. Confidential and Proprietary. © 2019Dr. UST Global Adnan Inc Masood 4 tl;dr – intelligent components lead to autonomous operations • Automation on Steroids • AI Trends in Automation • Message from the Trenches Confidential and Proprietary. © 2019 UST Global Inc 5 AI is an unstoppable force for autonomous operations Confidential and Proprietary. © 2019 UST Global Inc 6 As soon as it works, no one calls it A.I. anymore John Mcarthy Coined the term Artificial Intelligence in 1956 Confidential and Proprietary. © 2019 UST Global Inc 7 The AI Effect "A lot of cutting edge AI has filtered into "Every time we figure out a piece general applications, often without of it, it stops being magical; we being called AI because once something say, 'Oh, that’s just a becomes useful enough and common computation”’ enough it's not labelled AI anymore” Rodney Brooks Nick Bostrom "AI is whatever hasn't been done “As soon as it works, no one calls it AI yet.” - Tesler's Theorem as quoted any more” Douglas Hofstadter John Mcarthy Confidential and Proprietary. © 2019 UST Global Inc 8 Would this still be considered AI? Confidential and Proprietary. © 2019 UST Global Inc 9 Confidential and Proprietary. © 2019 UST Global Inc 10 Confidential and Proprietary. © 2019 UST Global Inc 2018 Automation Summit 11 Placeholder Confidential and Proprietary. © 2019 UST Global Inc 12 Confidential and Proprietary. © 2019 UST Global Inc 13 Enterprise AI Systems are built with recurring business processes which can be captured, and optimized. Confidential and Proprietary. © 2019 UST Global Inc 14 AI Building Blocks for an Enterprise Confidential and Proprietary. © 2019 UST Global Inc 15 If you're trying to understand Al’s near-term impact, don't think "sentience“ Instead think "automation on steroids” Andrew Ng Confidential and Proprietary. © 2019 UST Global Inc 16 UST SmartOpsTM Confidential and Proprietary. © 2019 UST Global Inc 17 UST SmartOps™ is an AI powered intelligent operations platform with 30+ capabilities Knowledge Intelligent Autonomous Automation Management Monitoring Operations Natural Language Business Transaction Autonomous Runbook Automation Understanding Monitoring Monitoring Data Fusion with Self Service Workflows User Behavior Tracking Anomaly Detection Computer Vision Cognitive RPA (OASIS) Knowledge Graph Support Activity Fault Prediction Robotic Process Application Preemptive Dark Data Mining Automation Performance Intervention Self Service/Virtual Cognitive Search Business Performance Self Healing Assistants Confidential and Proprietary.…© 2019 UST Global Inc … … … 18 Cognitive Search, Content Summarization Search Service Advanced Cognitive Search with and other paid/subscription content – understand Intent and Context Users Summarization Service Sales, and Marketing Teams – Cognitive Extractive and Abstractive Content Writers, Marketers, Search and Document Summarization Distributed Sales Team Summarization Corpora Internal Document subscription data Multi-media content, images, videos, infographics Confidential and Proprietary. © 2019 UST Global Inc 19 Scaling Cognitive Automation in The Enterprise How do I do AI & Automation like a tech giant? “[What are] industry best practices and “I have built 10 PoCs and I have no hope that innovations in the scaling of AI and Machine any of them will be put into production” Learning and connecting it to the business, to leadership, and to customers” “How do we automate some of our business functions more effectively? Right now there are “How do we benchmark ourselves against a lot of manual processes. How can we apply leaders in data science and automation like the autonomous operations concept to business Google, Uber, and Wayfair? processes? Confidential and Proprietary. © 2019 UST Global Inc 20 Enterprises face three key sets of AI & ML barriers Artificial Intelligence Usage Barriers – Survey of Decision Makers Opportunity costs of AI investment 24% There is not a unified belief in a clear ROI to AI 18% Don’t know what the We are not sure how to apply it to our business 16% right problems are We are not sure where to apply it to our business 15% Lack of skills to implement and operate such systems 22% Don’t know how 20% Change management to replace existing data and… to do it We don't have a well-curated collection of data to… 15% Introduction of new security threats 23% Privacy concerns 22% Workforce concerns about job security 21% Don’t trust it Concern about unintended, potentially negative, and… 18% Inability to maintain oversight and governance 17% ConfidentialSource: and2,594 Proprietary.Data and analytics ©decision 2019-makers UST whose Global firm is Inc interested in using/planning to use/currently usingAI - Forrester 21 Enterprise software companies building seamless AI into applications Confidential and Proprietary. © 2019 UST Global Inc 22 Applications ingrained with “AI/ML” features Schema generation & data Suggests next steps Natural language generation transformations ConfidentialSuggests and Proprietary. data© 2019 cleaning UST Global Inc Root cause analysis 23 EnterpriseAI Trends in Automation Confidential and Proprietary. © 2019 UST Global Inc 24 Measure what Matters - AI Life Cycle Management • Machine Learning Ops - MLOps • Ethical & Interpretable AI • Automated Machine Learning Getting Better with Intelligence • Generative Adversarial Networks • Transfer Learning & Pre-built Models • Reinforcement Learning • Augmented Intelligence Confidential and Proprietary. © 2019 UST Global Inc 25 Academia to Industry – The Pipeline MIT Computer Science and Artificial Intelligence Laboratory Clients share their problem statements and host sessions with researchers and UST to partner on focused research projects Access to researchers, open source tools, and visiting scholar program to engage, build, and learn with clients Collaboration with researchers and faculty to launch and manage strategic centers of excellence around specific business & research problems Confidential and Proprietary. © 2019 UST Global Inc 26 Machine Learning Ops - MLOps Confidential and Proprietary. © 2019 UST Global Inc 27 Data infrastructure must meet the unique requirements of AI Confidential and Proprietary. © 2019 UST Global Inc 28 AI infrastructure used to train ML models must be Confidentialavailable and Proprietary. © 2019 UST Globalin Inc perpetuity 29 Enterprise AI Development: Challenges Confidential and Proprietary. © 2019 UST Global Inc 30 The ML model building lifecycle is highly iterative and continuous Big data ingestion, processing and Identify data that is relevant to preparation the business goal Understand Integrate and enrich the Prepare data into an analytical data data set Development Measure the effectiveness Monitor of the model in ML the real world Models Production Run statistical and ML Use the model in Deploy Model algorithms to find the applications model Test the model to make Model scoring/ sure it will work Evaluate Model training inferencing in using ML applications algorithms Confidential and Proprietary. © 2019 UST Global Inc 31 Machine Learning Lifecycle Data Data Ingestion Data Analysis Data Validation Data Splitting Transformation Trainer Building a Model Model Validation Training At Scale Logging Roll-out Serving Monitoring Optimization Confidential and Proprietary. © 2019 UST Global Inc 32 Ethical & Interpretable AI Confidential and Proprietary. © 2019 UST Global Inc 33 2 Explain your AI Confidential and Proprietary. © 2019 UST Global Inc 34 “Responsible AI” foretells massive enterprise AI adoption Confidential and Proprietary. © 2019 UST Global Inc 35 Ethical Machine Learning and Interpretability Confidential and Proprietary. © 2019 UST Global Inc 36 Automated Machine Learning Confidential and Proprietary. © 2019 UST Global Inc 37 Auto-ML solutions automate key aspects of the model building lifecycle including feature engineering, algorithm selection, and tuning Confidential and Proprietary. © 2019 UST Global Inc 38 3 Data scientists aren’t just expensive. They are inefficient. Confidential and Proprietary. © 2019 UST Global Inc 39 Auto-ML platforms increase productivity by 1,000% Confidential
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages59 Page
-
File Size-