Cognitive IoT: What is Watson IoT? Amit Fisher Program Director, Product Offering Innovation, IBM Watson IoT Cognitive Offering Leader Member, IBM Industry Academy Email: amfisher@us.ibm.com 1
IoT is driving Digital Disruption of the Physical World Accelerating advances Are transforming every in technology part of business Improving operations Advanced analytics and lowering costs Cloud computing Creating new products Pervasive connectivity and business models Product Lifecycle Management Driving engagement and Embedded sensors customer experience 2
IoT represents enormous scale and impact 25 Billion Connected devices are growing at an exponential rate � Installed IoT Devices 2 by 2020 $3.6 Trillion Value derived from this connectivity drives massive monetary opportunity � Poten5al economic impact 1 per year by 2020 70% B2B Majority of value will be in B2B and B2B2C use cases � IoT value created in B2B use cases 3 3 Source: 1) McKinsey June 2015 2) Gartner November 2014
With IoT, companies are becoming more competitive with new ways to drive better business engagement How much energy cost can I reduce? Can I reduce maintenance costs by doing condition repairs instead of time-based maintenance How can I increase the utilization of my assets? We need to create multiple variants of How can my our products for different markets business How much more revenue can I generate from my current assets? deliver I need to reduce risk developing this complex device beIer client I need to reduce/eliminate my factory downtime due to unplanned outages experience I have assets deployed all over the place or beIer that need a repair process I need help managing complex outcomes development projects How can I best support our organization's with IoT environmental sustainability objectives? I have to prove we met regulatory specs to an auditor! I need to find new sources of revenue by moving to new service-business models I need to reduce the cost of running my building 4
With IoT, Clients are looking to… IoT Solutions ü Rapidly and securely connect devices ü Optimize operations Opera5ons Asset Performance Facili5es Mgmt Health & Safety Connected Products Work Mgmt Product Development ü Enable new business models Watson IoT Platform ü Engage with clients and markets in new ways Factories Facilities Vehicles Home Transport Health 5
IoT represents substantial market opportunity IoT Solutions 9.6% $88B Opera5ons Asset Performance 10.9% Facili5es Mgmt Health & Safety $55B Connected Products Work Mgmt Product $117B Development $70B Watson IoT Platform 22% $14B $5B $27B 12% $15B 2013 CAGR 2018 IBM Market Opportunity Factories Facilities Vehicles Home Transport Health 6
IoT Client Value Strategy Is your client looking to connect… Devices? Equipment? People? Start with the IoT Cloud Pla9orm Connect to… Secure connec5vity Connect Manage devices Store and archive data Informa5on Organize and transform Management Structure and unstructured Watson IoT Platform Real 5me Analy5cs Predic5ve Cogni5ve Data protec5on Risk Security analy5cs Factories Management Facilities Key and cert management Vehicles Home Transport Health 7
IoT Client Value Strategy Is your client looking to op<mize… Assets? Product Development? Safety? Start with the IoT Applica<ons Improve space u5liza5on Opera5ons Asset Performance Reduce energy usage Facili5es Mgmt Facility & Space Health & Safety Connected Products Work Mgmt Product Reduce 5me to value Development Improve lease mgmt Real Estate Op5mize resources Watson IoT Platform Product Development Increase ‘re-use’ Opera5onal risk Enable safety culture Health & Safety Life cycle mgmt Factories Configura5on mgmt Facilities Asset Management Vehicles Home Transport Health 8
IoT Client Value Strategy IoT Solutions Is your client looking to transform tradi<onal business with IoT… Invent new business models • Opera5ons Asset Performance Facili5es Mgmt Health & Safety Develop differen5ated • Connected Products Work Mgmt Product solu5ons Development Improve opera5onal efficiency • Watson IoT Platform Drive beIer customer • engagement U5lize IBM innova5on and a • Factories Facilities Consult to Run partnership Vehicles Home Transport Health 9
Cognitive Computing “Cognitive computing refers to systems that learn at scale, reason with purpose and interact with humans naturally . Rather than being explicitly programmed, they learn and reason from their interactions with us and from their experiences with their environment. ” John E. Kelly III 10
IBM Watson IoT Solution Solutions Applications Enabling new Optimizing business models operations for with integrated business impact solutions for industry Business Connecting Transformation Data via ecosystem Powered by and partner IBM Watson relationships Enabled Local by IBM Cloud Deployment Platform Everything you need to innovate with IoT 11
Cognitive IoT Cognitive IoT enables us to learn from, and infuse intelligence into, the physical world to transform business and enhance the human experience. 12
What is “Dark Data” ? Unstructured Data (a.k.a. “Dark Data” 80% of all WW data… Structured Data 20% of all WW data 13
How does it work? 14
How does it work? Ingestion – Indexes, Defining the Field of Defining Corpus of Curating the Content metadata and knowledge Knowledge Knowledge (Humans) graphs Training (via Machine Building a Reasoning Further training and fine Cognitive System Learning) Model tuning by user interaction 15
From Jeopardy! to APIs Natural Language Video and Processing (NLP) Image Analytics Machine Text Learning Analytics 16
Bluemix + Watson + IoT: a developers ‘candy store’ ! New Watson APIs That can apply to IoT Machine Natural language learning processing Video and image Text analytics analytics 17
Why IoT needs machine learning ? Why IoT needs machine learning Complex but Computable output output Inputs Inputs Things & Ensemble of Things Network of Things • Known laws govern system behavior • Mathematical equations become complex • A mathematical equation captures behavior • Computer algorithms can model system behavior System behavior (relationship between inputs and outputs) can be determined. Analytics models the equation or encodes the algorithm in software. 18
Why IoT needs machine learning ? as systems become more complex, generate more data, and integrate more data sources, we need machine learning to process & understand the extreme volumes of data output output output Inputs Inputs Inputs Large Network of Smart Things Interacting with Each Other Complex relations that are different under different contexts – and may be different at different times Scale, diversity and complexity make the Best option à determine correlations relationships between inputs and outputs between input and output to learn the hard to determine relationship Machine learning finds relationships between inputs and outputs — when it is hard to build a model. 19
Why IoT needs machine learning ? Inputs Inputs Inputs Increased Scale Increased Complexity Dynamicity, outputs outputs Complexity outputs Analyzed Model Computed Model Learned Model • “White Box” with quantifiable • “White Box” with computable • “Black Box” with learned relationship relationship relationship • Suitable for systems governed • Suitable for systems where • Suitable for large scale complex by hard laws of nature computational models can be networked systems with determined context-dependent and dynamically changing relationships 20
Case Study: Cognitive IoT for Pipeline Corrosion Prediction Machine Trained Learning Model Inspection Corrosion Data RMSE [mm/ Next Best Prediction Model year] ANSWERS Pipeline ~4x Yes/No? IoT Data IBM Pipeline IoT Data QUESTION Does the pipeline need Different Corrosion Models Machine-learned to be repaired/inspected? Model Business value of cognitive pipeline analysis Total corrosion-induced cost in the US pipeline sector is $8.6B/ year*. Improved corrosion predictions reduce maintenance and SPECIFIC GENERIC inspection costs . $12,000 corrosion-induced cost per mile of pipeline $2,000 per mile of inspection costs 2 million miles of pipelines world-wide Trained from fully Trained from whatever characterized pipes data is available * NACE Report 2015; National Association of Corrosion Engineers 21
Case Study: Open Cognitive Interface with Harman Example: Weather data for time and location ‘ Chances of stormy weather in Stores DB Detroit tomorrow is 20% ‘ Job site #123 is in Detroit MI WAV files over MQTT (Class, Location, Time) ( ‘ Tomorrow ’ ) WAV files over MQTT Question ’ s class (e.g. temp, rain, snow, wind Question as a string etc.) “ Will the storm hit job site #123 tomorrow ? ” Class= ‘ weather ’ Class= ‘ snow ’ 22
Recommend
More recommend
