Defense Information Systems Agency (DISA) officials say the challenge with artificial intelligence (AI) is not a lack of capabilities, but rather providing warfighters with valuable, AI-enabled, and easy-to-use capabilities.
Today at AFCEA NOVA’s Department of Defense Enterprise IT Day, Deepak Seth, AI technical lead for DISA’s Emerging Technologies division, said that while DISA has many capabilities to choose from in today’s AI environment, “the problem is… It’s about how you can utilize the functionality.” How can we drive them towards production?”
For example, large-scale language models (LLMs) provide users with models that are pre-trained on large datasets and can be fine-tuned for specific tasks. However, according to Seth, these models “tend to lack corporate knowledge.”
“So how do we [we] These pre-trained models can be extended with enterprise datasets, so when asked [the response] It’s based on data within the organization,” Seth said.
One of the functions DISA is working on is providing a question-and-answer model for government officials and warfighters.
“The idea is [to] We take all this information that we have and build it into some kind of application,” Seth said.
This feature is like this DISA’s Digital Conciergeexpected to be fully rolled out in August.
Concierge AI The idea is to integrate data with AI to minimize friction for users to find and analyze data. The digital concierge takes data from a controlled, non-sensitive setting and drops it into a database. Through the LLM, the database provides the answer to the user.
In addition to generative AI and LLM, DISA seeks to apply AI models, tools, and services to improve defensive cyber operations (DCO).
According to Seth, the challenge facing DCO analysts is not only that attacks are increasing, but that they are becoming faster and more sophisticated. The agency hopes to leverage AI tools to analyze incoming data and “detect anomalies” more quickly.
DISA released Request for Information (RFI) published on SAM.gov This report outlines that there is interest in exploring the potential applications of commercial AI/ML models, tools, services, and best practices to enhance and enhance current DCO capabilities and methodologies. did.
The agency has received several hundred responses and is currently reviewing them.