DARPA flags interest in potential AI research opportunity

The Defense Advanced Research Projects Agency has identified a potential new artificial intelligence exploration opportunity surrounding the combination and processing of sensor and environmental information to bolster machine learning.

The so-called “enabling confidence” effort, which DARPA flagged in a special notice last week, could become the latest AI research topic for the agency’s Microsystems Technology Office.

The listing, which signals officials’ interest in the area, states the effort aims to research whether input sensor and environmental covariance data could enhance the performance of ML systems. Processing those inputs accurately, it notes, “is the key enabler for optimally combining information originating from multiple heterogeneous sensors and subsystems.”

Covariance describes the relationship between two variables and how changes in one track with changes in the other.

If pursued, the notice says DARPA will approach the topic within the confines of an 18-month program, split evenly into two nine-month phases. Under the initial phase, those involved would work to assess the feasibility of achieving accurate covariance baselines through model experimentation, while the second phase would shift the focus to the potential of edge deployment.