In this talk I will present several examples of how AI models successfully inform interactive analytics and augmented intelligence solutions to transform national security focusing on cognitive, nuclear and climate security mission areas. I will start with cognitive security and discuss our approaches to detect, characterize, and defend against influence operations and disinformation campaigns based on novel in-house developed deep learning and natural language processing models. Our novel models capable of detecting and analyzing information narratives, understanding social media audiences, characterizing the information environment, and discovering causes and effects to enable precision communications. I will demo WatchOwl analytics (https://watchowl.pnnl.gov/) to analyze content and characterize audiences during COVID-19 infodemic.

Next, I present a suite of AI-powered analytics for nuclear security to detect, anticipate, and reason about proliferation expertise and capability development globally by learning from massive-scale unstructured dynamic real-world data. I will showcase our augmented intelligence tools that support nonproliferation and AI use cases. I will describe how we go beyond descriptive analytics and propose predictive and prescriptive analytics that feature graph neural networks to anticipate future collaboration patterns, authorship behavior, and capability evolution from dynamic heterogenous graphs, and ensemble models for causal discovery and inference to enable counterfactual reasoning about expertise and capability development. I will wrap up by discussing our new internal investment in AI that aspires to move beyond narrow AI and focus on rapid development and deployment of foundation models of knowledge for science and security applications e.g., climate security to augment domain scientists with the advanced ability to perceive and reason at scale previously unimagined.