Explore a comprehensive lecture on building trustworthy AI systems with common sense capabilities. Delve into the challenges and potential solutions for creating AI that can work effectively alongside humans. Learn about the CARE principles for responsible, adaptive, explainable, and collaborative AI. Discover how neuro-symbolic architectures can be used to incorporate common sense knowledge and reasoning into AI systems for tasks like story modeling and generation. Examine the speaker's vision for human-AI teams tackling important challenges in personalized medicine, cybersecurity, and investigative reporting. Gain insights into cutting-edge research on robust and explainable AI technology with real-world applications.
Overview
Syllabus
Introduction
AI in 2050
Lack of Commons
This Is Us
Lack of Common Sense
What is Common Sense
My Research Goal
Challenges
Knowledge Transfer
Explainable Collaborative
Framework
Question Answering
How Does AI Know
Common Sense Knowledge Graph
Knowledge Integration
Generating Questions
Train the Model
Constant Performance Improvement
The First Encouraging Step
Goal of Adaptivity
Summary
Task
Adaptive System
Argument Analytics
Summary and Vision
TaskOriented Dialogue AI
Detection of Malicious Content
Ideation Discovery
Technical Challenges
Devil Advocate
Linda Henry
Taught by
USC Information Sciences Institute
