Downloaded 46 times
Uploaded byAssist
Artificial Intelligence at Work - Assist Workshop 2016 - Nick Triantos - SRI
This document discusses the transformation of computing through artificial intelligence and machine learning. It summarizes that AI is becoming mainstream and reshaping the workplace, with machine learning being used for tasks like image recognition, data analysis, and speech recognition. Current virtual assistants from companies like Google, Amazon, Apple, and Facebook are discussed and compared. The document advocates for more specialized virtual assistants with domain-specific knowledge and personalized experiences based on understanding individual users.
More Related Content
Similar to Artificial Intelligence at Work - Assist Workshop 2016 - Nick Triantos - SRI
Artificial Intelligence at Work - Assist Workshop 2016 - Nick Triantos - SRI
- 1. POWERTHROUGH SIMPLICITY: THETRANSFORMATION OFCOMPUTINGTHROUGH ARTIFICIAL INTELLIGENCE NickTriantos, Managing Director, SRIVentures
- 2. About SRI Wearable Robotics TheInternet Computer Mouse Cybersecurity Siri Autonomous Robotics Space & Comms Diagnostic Ultrasound
- 3.
- 4. "AI Is AboutToGo Mainstream And ReshapeThe Workplace",TechCrunch, Sept 2015 AI-POWERED BUSINESS APPS
- 5. AI USE INTHE ENTERPRISE "2015State of Artificial Intelligence & Big Data in the Enterprise", NarrativeScience Machine Learning VPA Decision Support Reporting Analytics Robotics All of the Above 0% 6% 12% 18% 24%
- 6. WHAT MACHINE LEARNINGCAN DO Image Credit: Scikit-learn Categorization "What is this thing?" Supervised Categorization "Which known thing is this?" Unsupervised categorization "Which things are similar?" Data Analysis Regression "Predict a numeric result" Dimensionality Reduction "What's important here?"
- 7.
- 8. APPROACHING "GOOD ENOUGH" "KPCBINTERNETTRENDS 2016 ", Mary Meeker, Kleiner Perkins, p. 117
- 9. SPEECH RECOGNITIONTODAY "KPCB INTERNETTRENDS2016 ", Mary Meeker, Kleiner Perkins, p. 119
- 10. USER INTERFACE ISSTILL USEFUL
- 11. VOICE: MIXED BLESSING TheGood • Access any part of user interface instantly • Easy and convenient • Adaptable & Personalized The Less Good • Can fail with connectivity loss • Some security challenges • Harder to create expressive user interface
- 12. OVERVIEW OF CURRENT VIRTUALASSISTANTS
- 13. GOOGLE ASSISTANT • GoogleHome & Google Allo • Context-aware responses • Leverages the other data Google knows about you • Only available through Google apps so far "This is not a thing that's going to figure itself out in 2016. It's going to be a long journey." – Mario Queiroz,VP Product, Google
- 14. AMAZON ECHO • Funcloud-based platform • Excellent hardware for far-field voice recognition • Excellent ASR and getting even better • Extensible, and now appearing in non-Amazon products • Very limited capability to have a conversation • Requires connectivity
- 15. APPLE SIRI • Siricoming to MacOS • Limited conversational context • Opening Siri to developers (at last)
- 16. FACEBOOK "M" • 100%human-powered today • Goal to be 100% machine-driven in 5 years • Access to huge database of your interests and those of your friends • Heavy overlap with core ad serving technologies
- 17.
- 18.
- 20. DEEP KNOWLEDGE & THE RISEOFVIRTUAL SPECIALISTS
- 21. SRIVIRTUAL ASSISTANT ("VPA") •Deeply conversational • Well-suited as "Virtual Specialist" • Custom speech recognition models • Highly customizable ontologies • Embedded in numerous platforms • Can work without the cloud
- 22. MORE NATURAL COMMUNICATION FLOW Transfer €200to checking. From which account? How does my savings look? Your savings balance is €2.200. Should I continue the transfer? Will there be any fees? No, your savings account has unlimited free transfers OK, go ahead and transfer the money. Your €200 transfer from savings to checking is complete.
- 23. DOMAIN SPECIFIC KNOWLEDGE Transfer FromTo Account Savings Checking Restrictions Restrictions Balance Warnings ! Fee warning Balance warning!
- 24.
- 25.
- 26. SOCIAL INFLUENCES • Opaqueadaptable recommendations • Leverage huge pool of data in social networks • Likely to dramatically improve over time • Allow for high-value advertising if done well
- 27.
- 28. Virtual Assistant Platform PERSONALIZATIONHUB Present a better experience, personalized to the user, based on: • Who the user is • What the user usually likes • What state the user is in • Where and when the interaction with the user occurs Context aware Emotion detection Preference learning Biometric ID Application
- 29.
- 30.
- 32. YouTube: Meskel Square,AddisAbeba, Ethopia https://www.youtube.com/watch?v=UEIn8GJIg0E YouTube:Autonomous Intersection Management, UT Austin https://www.youtube.com/watch?v=4pbAI40dK0A
- 34. CLOSING THOUGHTS • Cheap sensors+ machine learning = feasible AI • Early tools are rough, but rapidly improving • Virtual specialists & AI will be able to augment and later replace some jobs • Collect data now, learn from it quickly, or be passed by those who do
- 35.