News

A new research trend? DARPA wants to fund XAI (Explainable Artificial Intelligence)

Published: 18 August 2016

Artificial Intelligence (AI) is becoming more and more integrated in our daily lives: AI agents might decide if you get a bank loan, or if your job application will ever reach human eyes. Not everyone is comfortable with this trend, since we don't always know exactly how the AI comes to its decision. AI learns from existing data to predict future data, but its inner workings can be a mystery even to the AI's programmers. That's a problem if the AI is making life or death decisions, as it would in missile systems or unmanned drones. In response, the Defence Advanced Research Projects Agency (DARPA) is seeking research in Explainable AI, that is, AI that will output an answer and the criteria that led the system to its decision. Here is an example from DARPA's slides: instead of simply classifying a photo as one of a Kentucky warbler, it would output something like "this is a Kentucky warbler because this is a yellow bird with a black cheek patch and a black crown."

DARPA launched its XAI project August 10, 2016. Abstracts for potential projects are due September 1. The winners will not be receiving grants but will rather be receiving procurement contracts and cooperative agreements. You can read the full details of the initiative, or a short article summarizing it.

 

Back to top