InTTrust -IBM Artificial Intelligence Event
- 1. A Hitchhiker’s Guide to AI The IBM Highway Michail Pagiatakis Cloud & AI Solutions BDM InTTrust S.A
- 2. AGENDA AI Business Unusual42 Infra Does it Matter? Automatic for the People IBM A Powerful Ally Pride & Prejudice
- 3. 42 The answer to life the universe and everything
- 4. Self driving cars react to changing conditions Waze provides a personalized driving experience Uber delivers food that you like and is the right temperature Netflix provides personalized recommendations AI is Everywhere Influencing Everything We Do Some Very successfully Some not as successful But ALL based on DATA and AI 1. Business Value
- 5. Chilean forestry companies (two examples) Timber harvesting $20M/year + 30% fewer trucks UPS Air network design $40M/year + 10% fewer planes South African National Defence Forces Force and equipment planning $1.1B/year Motorola Procurement management $100M–150M/year Samsung Electronics Semiconductor manufacturing 50% reduction in cycle times SNCF (French rail network) Scheduling and pricing $16M/year rev + 2% lower operating expense Continental Airlines Crew rescheduling $40M/year AT&T Network recovery 35% reduction spare capacity Grantham, Mayo, Van Otterloo & Co. (GMO) Portfolio optimization $4M/year Source: Edelman Finalists, http://www.informs.org or http://www.scienceofbetter.org Delivering massive return on investment 5
- 6. A transportation company used analytics to optimize route planning every 10 minutes The company saved millions of dollars annually by eliminating miles of unnecessary driving. Manage hundreds of constraints on tankers, drivers and cargos to increase profits and ensure safety With an IBM ILOG CPLEX Optimization Studio solution, the company found the optimal combination of drivers, trailers and tank- washes across its whole logistics network. C H A L L E N G E : S O L U T I O N : 6
- 7. FAKE mix and match your algorithms built on- premises with those in the cloud
- 9. PowerAI Vision: ”Point-and-Click” AI for images & video Label Image or Video Data Auto-Train AI Model (no coding, just point and click) Package & Deploy AI Model
- 10. PowerAI Vision core capabilities
- 11. Upload video and select Label Objects Pick which frames you want to label objects in or let PowerAI Vision automatically capture frames on an interval you specify
- 12. Label the frames with object names Using a subset of the video frames, label the objects with bounding boxes exactly as you would for regular image object detection
- 13. Build a simple model and then use it to auto-label more frames
- 14. My original tagged frames Auto labelling helps to label more frames I just need to add in other labels it missed and then retrain with this additional data for improved accuracy
- 15. Test your model – upload a video and click “Detect”
- 17. that in turn requires much higher computation to train as neural networks go deeper, they provide a dramatic increase in accuracy.
- 18. requires different sets of skills FINE-TUNE & DEPLOY experience all that pain again and iterate MAINTAIN ACCURACY iterate faster and do it againassisted parameter selection and tuning ~80% of an AI project’s is time spent here DATA PREPARATION up and running over a quick lunch time spent drops from 80% to 30% extremely long training times curtailing broader proliferation BUILD, TRAIN, OPTIMIZE 9 days to train a model becomes 4 hours weeks to months UP & RUNNING imagine if everyday users could contribute business domain expertise, help with data preparation, and even build initial models so data science teams could focus on fine tuning the models data science skill needed as the hard stuff happens here - help with quick detection of sub-optimal hyper parameter selection, ‘what if’ exploration ... this is where you want data science teams to spend their time
- 19. hardware software Watson Machine Learning Accelerator co-optimization + open innovation
- 20. data preparation score model deploy model build, train, and optimize model ingest new data, re-score and deploy new model up and running $$$ monetize infer model maintain model accuracy
- 21. 21 “likely to be one of the most attractive platforms in the future, modern, open, flexible, and suitable for a range of users, from expert data scientists to business users.” 2017 reddot Design Award Winner IBM Watson Studio Local
- 22. 22 Case for AI Automation: AI Workflow’s Bigger & More Complex Majority spent on data wrangling! Ground Truth Gathering Data Cleansing Feature Engineering Model Selection Parameter Optimization Ensemble Model Validation Model Deployment Runtime Monitoring Model Improvement Source: https://www.kaggle.com/paultimothymooney/2018-kaggle-machine-learning-data-science-survey AI Lifecycle Management
- 23. IBM’s Strategy for Automation of AI Development Transfer Learning • Transfer knowledge learning in one deep learning system to apply to a different domain • Featured in Watson Services, available through Watson Studio Neural Network Search • Just bring data and automatically generate a custom deep neural network through searching the best architectures for the input data • NeuNetS as a feature of Watson Studio AutoAI Experiments | Pipeline optimization • Auto clean data, engineer features, and complete HPO to find the optimal end to end pipeline New
- 24. Watson OpenScale along withWatson Studio andWML enables enterprises to operationalize AI across the enterprise 24 IBM TOOLS Data Scientist App Developer Build AI RunAI 3RD PARTY IDE & FRAMEWORKS IBM AI RUNTIME Watson OpenScale Automated Anomaly and Drift detection Business KPIs Watson Studio Watson Machine Learning Manage AI at scale 3RD PARTY RUNTIMES Build Deploy and run Operate trusted AI Business user Consume AI Fairness and Explainability Inputs for Continuous Evlolution Accuracy Validation and Feedback SPSS Modeler Custom (Kubernetes etc.) Microsoft Azure ML Amazon Web Services Keras Pytorch Scikit-learn Spark ML Caffe2 … Watson Knowledge Catalog Data Profiling Quality and Lineage Data Governance Organize and Govern data Data Engineer Organize Data for AI
- 25. IBM Watson OpenScale Automate & Operate AI at Scale Production monitoring for compliance Detect and mitigate model bias; audit and explain model decisions Ensure models resiliency in changing situations Detect drift in data and anomaly in model behavior; specify inputs and triggers to model lifecycle Align model performance with business outcomes Correlate model metrics and business KPIs to measure business impact; actionable metrics and alerts
- 26. IBM Watson & Cloud © 2019 IBM Corporation *** IBM & Business Partner Only *** 26 Business stakeholders do not trust AI. 60% of companies see regulatory constraints as a barrier to implementing AI. – IBM IBV AI 2018 63% cite availability of technical skills as a challenge to implementation. - IBM IBV AI 2018 Without expensive Data Science resources handholding multiple AI models in a production application: 1. No way to validate if AI models are compliant with regulations and will achieve expected business outcomes before deploying 2. Difficult to track and measure indicators of business success in production 3. Resource intensive and unreliable processes for ongoing business monitoring and compliance 4. Impossible for business users to feedback subtle domain knowledge into model lifecycle
- 27. IBM Watson & Cloud © 2019 IBM Corporation *** IBM & Business Partner Only *** 27 Why worry about AI and compliance? A leading bank was fined $175M for charging higher fees and interest rate to more than 30,000 minority borrowers. A leading credit card company was fined $75M for selling products to Spanish speaking customers at a much higher interest rate compared to others. http://www.fsb.org/wp- content/uploads/140407.pdf“… 50%of our bonuses are linked to how we manage risk …” – VP, AI platforms, Large American bank
- 28. Align model performance with business outcomes Correlate model metrics and business KPIs to measure business impact Actionable metrics and alerts Ensure that models are resilient to changing situations Detect drift in data and anomaly in model behavior Specific inputs and triggers to model lifecycle Production monitoring for compliance and safeguards Detect and mitigate model biases Audit and Explain model decisions Model Validation and acceptance Watson OpenScale will help validate and monitor AI models, deployed anywhere, to help comply with regulations and mitigate business risk Foundational to all AI implementations Required in regulated industries and use cases – FSS, HR etc. in short term; others longer term* Required to meet transformational goals * E.g. Fair lending practices in finance vs. GDPR across all industries Current capability Upcoming capability
- 29. Deploy enterprise AI anywhere, on the cloud(s) of your choice Unify AI across private, hybrid, and multi-cloud landscapes Align AI services and workflows with the data they rely on Tap into the open innovation of Cloud Paks and OpenShift 29 Watson Anywhere One Platform, Any Cloud Cloud-native container platform and operational services Cloud Pak for Data A one of a kind, pre-integrated set of data and AI services delivered within an open and extensive cloud native platform Hyperconverged Private Cloud System Organize Data Analyze DataCollect Data Infuse AI Everything you need for enterprise AI
- 30. 30 Watson Studio Watson Machine Learning Watson OpenScale Watson Knowledge Catalog Data Profiling & Prep Quality & Lineage Policy-based Governance Visual Design Develop & Train Lifecycle Mgmt Run & Optimize Model-ops Dynamic Retraining KPIs & Accuracy Explainability& Lineage AutomatedOptimization Prepare and Organize Data Build and Train AI Models Deploy and Run AI Models Manage and Operate Trusted AI One Unified Experience Watson AI Tool Suite A modular set of tools for creating and operationalizing custom AI models AutoAI Lifecycle Automation – “AI generating AI”
- 31. Watson APIs • Speech and language • Tone and visual recognition • Empathy and personality • Behavioral Insights • Conversational Build interactive AI applications, processes and product experiences accessible from any device Swift Watson Studio Supported by an active Github expert community
- 32. Watson Applications Speed time-to-value with pre-built AI applications for common use cases Watson Assistant Watson Discovery 32 Business Automation Financial Crimes Business Analytics Case Management Robotic Automation Anti-money Laundering Payments & InsurancePlanning & Budgeting Business Intelligence Sales Forecasting Customer Cross-sell Conduct Surveillance Strategic Partnerships Watson Health Compare & ComplyExpert Assistance Voice of the Customer Next-gen Call Center Knowledge Work Customer Experience Payer / Provider Care & Benefits Patient Experience * * *