DARPA's Explainable Artificial Intelligence - XAI Program

Explore DARPA's Explainable Artificial Intelligence (XAI) Program in this keynote address from the IUI2019 conference. Delve into the importance of XAI, supervised machine learning, and the program's strategy and goals. Learn about various XAI techniques, including heatmaps, black box analysis, network dissection, and language translation. Discover applications in autonomy, differentiable physics, and deep learning. Examine surprising results, program updates, and engage in a Q&A session covering topics such as data analytics versus autonomy. Gain insights into the challenges and progress in developing AI systems that can explain their decision-making processes, enhancing trust and understanding between humans and artificial intelligence.

Intro
DARPA Explainable AI
Why XAI
Supervised Machine Learning
XAI Strategy
XAI Program Goals
XAI Program Structure
Heatmaps
Black Box
Network Dissection
Language Translation
Gann
Autonomy
Differentiable Physics
Surprising Results
Deep Learning
Examples
Results
Program update
QA
Data Analytics vs Autonomy
Discussion
correctness of explanations