Few technologies are more consequential than artificial intelligence (AI) and machine learning. Exploring the bridge between the two for ways to improve homeland security operations is the focus of a new Department of Homeland Security Science and Technology Directorate (DHS S&T) Community of Interest (COI).
AI can, for example, help responders and law enforcement prioritize large amounts of data from a variety of places to expedite decision making.
“Eventually data becomes so large, no person can understand it or use it effectively,” said Dr. Patrick Carrick, Chief Scientist at DHS S&T. “It takes automated techniques to do that.”
Carrick recently assembled the COI for anyone in DHS with an interest in the subject. DHS operational components are expected to be widely represented in the community given the relevance to the homeland security mission.
Hundreds of personnel attended the kickoff meeting, both online and in person, which spoke volumes about the expected threats and opportunities associated with AI and machine learning.
The event began with Dr. Carrick providing an overview of the “state of the art” in AI/ML, describing its wide-ranging implications for the department, and detailing plans for the future of the community. This was followed by questions and comments from various components, experts and non-experts alike, who recognized they would each play a role in harnessing the power of AI.
The event continued with talks from other members of S&T, featuring a demonstration of an open source AI development platform followed by a discussion of the relationship between AI and the human mind.
“Not all AI/ML is relevant to everyone and what they’re doing,” said Carrick, citing perspectives on AI/ML that may deserve their own conversations, such as finance or transportation safety. “There is so much going on across so many different sectors. We want to make it so those with shared interests can go off and work on things together, then share their findings with the group and across the entire department.”
This discussion serves as a prototype for the future of the COI, with the goal of sharing valuable AI/ML information and insights from around the department. Participants are optimistic about the future of the COI and the DHS study of AI/ML. There was some discussion about eventually sub-dividing into even smaller mission-specific or interest area groups to facilitate the deeper exploration of AI/ML.
Other departments, like the Department of Defense (DoD), have similar communities that Carrick hopes will someday be able to exchange knowledge with the budding DHS COI. Carrick hopes to start COIs for Quantum Computing and Synthetic Biology as well, both of which are among the federal government’s highest R&D priorities.
How AI is Transforming Emergency Response Systems to Help Save Lives
Corti is a Copenhagen-based artificial intelligence platform that uses machine learning to provide emergency medical personnel with life-saving diagnostic assistance. The system analyzes voice patterns over the phone to help emergency personnel recognize typical patient symptoms. Medium
Can Artificial Intelligence Help U.S. SOCOM Track Weapons of Mass Destruction?
Countering WMDs a “daunting global problem” that requires advanced capabilities for analyzing intelligence from multiple sources — a task ready made for the industry’s newest machine learning tools that train computers to find things and people. There are now 75 National Geospatial Intelligence Agency analysts working full time at SOCOM headquarters in Tampa. Space News
Artificial Intelligence for Disaster Relief: A Primer
Some of the most-actionable information during a crisis comes from citizens, social media users, and on-the-ground aid workers. This info can help rescue workers find more victims in less time. Larger-scale behavior and movement data, run through predictive machine learning models, can help officials distribute supplies to where people are going, rather than where they were. Lexalytics
Why Information and Intelligence are Vital in Disaster Relief
These tools were used in a real-world situation following the 2015 earthquakes in Nepal, where Rescue Global and academics from the Orchid Project took pre and post-disaster imagery, utilised crowd-sourced data analysis and machine learning, to identify locations affected by the quakes that had not yet been assessed or received aid. Rescue Global
Comparison of Word Embeddings and Sentence Encodings as Generalized Representations for Crisis Tweet Classification Tasks
This paper explores the idea of building “generalized” classifiers for filtering crisis tweets, classifiers which can be pre-trained and ready to use in real-time using a simple feature-based adaptation approach, where tweets are represented as dense numeric vectors of reduced dimensionality using either word embeddings or sentence encodings. Tweet representations based on GloVe embeddings produced better results than the representations that use other embeddings, when employed with traditional supervised learning algorithms University of Illinois at Chicago (.pdf)