Walking Through The Aisle of Artificial Intelligence
- 1. Sridhar Seshadri Jul 2017 Concepts, Industry Trends & Real World Applications Walking Through The Aisle of Artificial Intelligence
- 2. Hi there! A quick intro… • Earned certification from Wharton and Johns Hopkins on Product Design Thinking and Data Science • Sweated at RGTU & BITS,PILANI for earning BE(Honors) and MS • Coaching PG students and Researchers from age of 17 with practical applicability of technology • Expert Architect of SMAC (Social, Mobile, Analytics and Cloud) Technologies Currently, Director of Technology & Products for Austin based start-up in AI space funded $50 Million by IBM, Microsoft Venture, Intel, Norwest Capital, USSA and is positioned 8th on the Forbes list of top 50 AI Startups. Learner, poet, author, technologist, entrepreneur, a music lover, and an avid photography explorer.
- 3. Agenda 1. Window to the World of AI 2. DIY Intelligent Alarm Clock! 3. Automation vs Intelligence 4. What is AI? 5. What contributes to AI? 6. Intelligence to a Machine? 7. AI Toolkit 8. Four Phases of AI System 9. Match of Golf 10. NLG 11. AI Robot Design Spec. 12. Mobile Game AI Use Case
- 4. AI in the trend
- 5. Window to the world of AI Focus: Big Data Analytics Focus: Data Science Focus: Enterprise AI Focus: Industry Transformation Focus: Structured Data TRADITIONAL ANALYTICS BIG DATA PLATFORMS MACHINE LEARNING PLATFORMS AUGMENTED INTELLIGENCE PLATFORMS INDUSTRY COGNITIVE CLOUDS Structured Data Unstructured Data AI Technical Services e.g. Watson AI/ML Tools e.g. Azure ML User Engagement Process Intelligence Cloud scale New Business Models 9
- 6. When (Room temp = 25 Degree C), Action = turn on the Air Conditioner Based on body heat pattern regulate the temperature of the Air Conditioner What is AI - Quiz
- 7. Alarm Clock Automatic Gun How will you make these intelligent or smart?
- 8. Signals (Metrics) 1) Sleep Cycle (when do you get up and when do you sleep) 2) Snoring Noise Pattern 3) Outdoor activity 4) Health activity monitoring 5) Calendar Schedule Based Prompts Etc.. Intelligent Alarm Clock
- 9. Signals (Metrics) 1) Biometric Pattern Matching 2) Pressure Matching 3) Modes based on Person’s profile ex. Taser, Pellet, Bullet etc. Etc.. SMART GUN
- 10. Automation • A task that has a fixed process and can be accomplished by following the same steps every time • It is based on conditions (ex. IF THEN ELSE) • Occurs at a fix frequency Intelligence • No fixed process to accomplish, is critical thinking problem statement • Is beyond a IF THEN ELSE, and is mostly probabilistic • Based on a complex pattern and needs analytical way to think through Automation vs Intelligence
- 11. Design a Intelligent Bullet Proof Vest
- 12. What is AI? Mr.Wilson: “But Mr. Holmes, how did you know I had been in China?” Sherlock Holmes: The fish that you have tattooed immediately above your right wrist could only have been done in China. That trick of staining the fishes’ scales of a delicate pink is quite peculiar to that country. When, in addition, I see a Chinese coin hanging from your watch‐chain, the matter becomes even more simple.” Mr. Jabez Wilson laughed heavily….. “Well, I never!” said he. “I thought at first that you had done something clever, but I see that there was nothing in it after all.” As soon as it works, no one calls it AI anymore.” — John McCarthy The Red‐Headed League, Arthur Conan Doyle Conversation between Mr. Jabez Wilson & Mr.Holmes
- 13. What Contributes to AI? Reasoning (Inductive, Deductive) Learning (Auditory, Episodic, Motor, Observational, Relational, Spatial) Problem Solving (decision based, critical thinking) Perception(acquire, enrich, filter, organizing sensory data) Linguistic (Comprehend from speech, tone, written and verbal lang.)
- 14. - 5 Senses - Taste, Sight, Smell, Hearing and Touch - Logic - Rational & critical thinking - Ability to think on the feet - Action - Execute on the best option of all - Measure impact - Get better with execution next time What is Intelligence to a Machine? “ Machine shall mimic alike a human baby”
- 15. Toolkit AI •Machine Learning •Natural Language •Computer Vision •Gesture Control •Dialogue & Conversation
- 16. Four Phases of AI System Act on the advice thru software/hardware Learn from feedback, models, and advice that is accepted / refuted Connect to the domain data Discover right algorithm Define the relevant use case and domain context Model the logic to solve for the use case Advise based on the outcome of the algorithm Assure with the confidence score 8 Define & Model Connect & Discover Advise & Assure Act & Learn
- 17. AI – Match of Golf Problem Statement: “Will we be able to play a match of golf tomorrow looking at the weather condition (Outlook = Rainy, Temp = Mild, Humidity = Normal, Windy = True)?” Posterior probability of occurrence of event What do we need to know? Factors on which the probability of occurrence and non occurrence depends.
- 18. AI – Match of Golf Phase 1 : Define & Model • Define the metrics that shall impact the final outcome • Collect the data • Enrich the data • Choose the right model • We need a math model for calculating posterior probability • Transform the freq. data to likelihood data and finally use Naïve Bayesian Equation to calculate the Posterior Probability for each classification (occur and non-occur)
- 19. AI – Match of Golf Phase 2 : Connect & Discover •We need a math model for calculating posterior probability •Transform the freq. tables to likelihood tables and finally use Naïve Bayesian Equation to calculate the Posterior Probability for each classification (occur and non-occur)
- 20. AI – Match of Golf – Connect & Discover Outlook Temp Humidity Windy Play Golf Rainy Hot High FALSE No Rainy Hot High TRUE No Overcast Hot High FALSE Yes Sunny Mild High FALSE Yes Sunny Cool Normal FALSE Yes Sunny Cool Normal TRUE No Overcast Cool Normal TRUE Yes Rainy Mild High FALSE No Rainy Cool Normal FALSE Yes Sunny Mild Normal FALSE Yes Rainy Mild Normal TRUE Yes Overcast Mild High TRUE Yes Overcast Hot Normal FALSE Yes Sunny Mild High TRUE No Define: Signals/Metrics Outlook, Temp, Humidity, Windy, Historic Outcome Play Golf Apply the chosen model on the data Probability of Yes = 9/14 No = 5/14 Frequency Table Play Golf Frequency Table Play Golf Yes No Yes No Outlook Sunny 3 (3/9) 2 (2/5) Temp. Hot 2 (2/9) 2 (2/5) Overcast 4 (4/9) 0 Mild 4 (4/9) 2 (2/5) Rainy 2 (2/9) 3 (3/5) Cool 3 (3/9) 1 (1/5) 9 5 9 5 Frequency Table Play Golf Frequency Table Play Golf Yes No Yes No Humidity High 3 (3/9) 4 (4/5) Windy FALSE 6 (6/9) 2 (2/5) Normal 6 (6/9) 1 (1/5) TRUE 3 (3/9) 3 (3/5) 9 5 9 5 Historic Behavior
- 21. AI – Match of Golf – Outcome Prediction for a day, where Outlook = Rainy Temp = Mild Humidity = Normal Windy = True Likelihood of Yes = P(Outlook=Rainy|Yes)*P(Temp=Mild|Yes)*P(Humidity=Normal|Yes)*P(Windy=True|Yes)*P(Yes) = 2/9 * 4/9 * 6/9 * 9/14 = 0.014109347 Likelihood of No = P(Outlook=Rainy|No) * P (Temp=Mild|No)*P(Humidity=Normal|No)*P(Windy=True|No)*P(No) = 3/5 * 2/5 * 1/5 * 3/5 * 5/14 = 0.010285714 Normalize P(Yes) = 0.014109347/(0.014109347+0.010285714) = 0.578368999 P(No) = 0.010285714/ (0.014109347+0.010285714) = 0.421631001
- 22. AI – Match of Golf – Outcome Phase 3: Advice & Assure “It’s is likely that you can play the game of Golf if the weather conditions are, Outlook = Rainy, Temp = Mild, Humidity = Normal, Windy = True. Assurance : Confidence Level for The Prediction = 57% Phase 4: Act & Learn Add the Weather condition and the outcome back to the original database
- 23. NLG - NLP • Natural Language Generation (NLG), a subfield of artificial intelligence (AI) which produces language as output on the basis of data input, is not a new concept • There are a plethora of ways the technology is being employed, primarily to improve human productivity, customer engagement and operational efficiency
- 24. Skills: Stereoscopic vision, Speech Recognition System, Speech Synthesis System, Fall & Crash Proof. HEAD: Ears: 2 Omnidirectional microphones Eyes: 2 Webcams HiDef Movements: 4 Servos (Up-Down & Left-Right) Mouth: 20 Leds Nose: 1 Led Connection: Wifi Pcb 802g & antenna Controller: 1QPr2 pcb controller BODY: Sensors: 4 Ultrasonic sensors, 1 Sharp Infrared sensor, 3 Generic Infrared sensor Motor: 2DC Motors with Magnetic Encoder (170 RPM) Wheels : 2 Wheels (rear) + 1 Free Wheel (front) Sound 2 High Quality Speakers Controllers: 1 Mini-ltx main board powered by ATOM & Nvida ION Graphics 1 QPr1 pcb controller Battery: 7.5Ah one battery Robot Status: 1 Lcd Display 20x4 Artificial Intelligence Robot Design
- 25. Spender Profile SESSIONS/DAY SESSION TIME AGE IN GAME LEVEL DEVICE 4 85 SECS 2 DAYS 8 IPHONE 6 HISTORICAL PLAYERS CURRENT PLAYERS Mobile Game Company - Use Case Identify historical churners Determine play patterns Identify key churn indicators Create multiple profiles Search players w/ similar profiles Make predictions
- 26. POTENTIAL CHURNERS AI SDK automatically messages predicted churned users Automatically flight incentives Track ‘rewardees’ for 14 days AWESOMENESS! You’re doing great, greatness deserves rewarding! TOOL AGNOSTIC Direct to game client Through game server Through marketing tools AI SDK has its own tool as well Player Experience
- 27. Sridhar Seshadri Email: sridhar.seshadri@outlook.com Linkedin: linkedin.com/in/seshadrisridhar/ Thanks for listening through patiently! Questions