Jacob Andreas - Natural Language Explanations of Deep Networks

Explore cutting-edge research on natural language explanations of deep networks in this seminar presented by Jacob Andreas at MIT. Delve into various explanation strategies, including medical diagnosis explanations, multiagent communication, and meaning representations. Learn about translation evaluation techniques, multiagent system explanations, and abstraction methods. Examine neuron descriptions, natural language processing, and conversational representations. Gain insights into the challenges and future directions of this field as the seminar concludes with a comprehensive summary and thought-provoking discussion.

Intro
Explanation Strategies
Medical Diagnosis
Explanations
Multiagent Communication
Meaning Representations
Translation Evaluation
Multiagent System
Explanation Techniques
Abstractions
Neuron Descriptions
Natural Language Processing
Natural Language Labels
Conversational Representation
Summary
Challenges
Conclusion