DARPA, AI, and the XAI Program

   August 22, 2018, 4PM

302-308

 

Abstract:

In this talk I will provide an overview of DARPA’s history in AI, summarize DARPA’s current efforts in AI, and talk in more detail about the Explainable AI (XAI) program. The goal of XAI is to create a suite of new or modified ML techniques that produce explainable models that, when combined with effective explanation techniques, enable end users to understand, appropriately trust and effectively manage the emerging generation of AI systems. Eleven development teams are creating new ML techniques to produce more explainable models and also developing new human-computer interaction techniques to generate effective explanations. XAI is finishing the first year of the four-year program. By November 2018, the eleven teams of will have conducted initial evaluations to determine how well XAI explanations improve user understanding and trust of XAI systems, as well as user task performance when using XAI systems.