On National Security | Analyzing intelligence in the age of ChatGPT

by

Artificial intelligence had a huge moment in 2022. The chatbot ChatGPT, developed by OpenAI, gained notoriety for its ability to engage in seemingly human-like conversations, sparking curiosity and serious conversations about where this technology is headed.

Applications in national security and space are poised to benefit from this new age of AI, says technologist Patrick Biltgen, principal at the defense and intelligence contractor Booz Allen Hamilton. He says the industry is just beginning to grasp the potential of so-called generative AI, with tools like ChatGPT that create coherent and convincing written content and models like DALL-E 2 that come up with realistic images from a description in natural language.

Defense and aerospace organizations have long sought AI for its ability to automate tasks, shorten decision cycles and bring autonomy into systems. “But after ChatGPT took the world by storm, a lot of people are asking: How can this help my mission?” Biltgen said.

One of those missions could be space domain awareness, where AI can help to analyze objects in space and, more importantly to military leaders, determine the intent of maneuvering satellites. Human analysts today make judgment calls on whether an object approaching another object has hostile intent. Biltgen says an AI model could be trained to provide advice and “cue an analyst or an operator into a range of possibilities.”

This type of predictive analysis is harder than it sounds because hostile attacks in space “don’t happen very often,” he says, and there is a limited amount of physics-based data to train the models. “The trickiest thing is trying to model human intent.”

Russia’s invasion of Ukraine is a perfect case in point. “The invasion looks totally obvious in hindsight, but when they were building up forces and equipment in February, I and many others thought it was a bluff, and I was wrong,” says Biltgen. “We didn’t know what Putin really intended.”

Satellite maneuvers in orbit mostly look benign, but adversaries will keep testing the limits. “This is a very well known military tactic,” he says. “You fly right up to the edge of the other person’s country. You fly right along the border. You go through the international waters. And I think you’re seeing some of that in space where many operators have normalized the ability to move.”

For intelligence analysts trying to predict a hostile act in space or on Earth, generative AI could be game-changing if models are adequately trained.

The GPT chatbot, which stands for Generative Pre-Trained Transformer, was trained on a general body of knowledge and natural language processing. A GPT for national security analysts, for example, would be pre-trained “with all the intelligence reports that have ever been written, plus all of the news articles and all of Wikipedia,” Biltgen says.

So will AI put intelligence analysts out of work? Biltgen doesn’t think so, at least not for now. Former director of the National Geospatial-Intelligence Agency Robert Cardillo years ago predicted that bots would soon be analyzing most of the imagery collected by satellites and replace many human analysts, but that vision has not yet materialized.

A lot of AI-aided reporting today is very formulaic and not as credible as human analysis, he adds. “Intel analysts are building upon their knowledge of what they have seen happen over time.” But it is conceivable that an algorithm could be trained for activity forecasting, which will be “really hard to do because human life and geopolitics is very messy.”

Biltgen’s final analysis: “I don’t believe you can make a predictor machine, but it might be possible for a chatbot to give me a list of the most likely possible next steps that would happen as a result of his series of events.”

And what does ChatGPT have to say about this?

“With the ability to analyze vast amounts of data, detect patterns and anomalies, and make predictions and decisions at a speed and scale that humans are unable to match, AI can help to identify and thwart threats before they occur, improving the effectiveness and efficiency of national security operations. As such, it is likely that AI will play an increasingly important role in national security in the coming years, and its adoption and development will be a key priority for many governments around the world.”

OK, if you say so.


Sandra Erwin

 

Sandra Erwin covers military space for SpaceNews. She is a veteran national security journalist and former editor of National Defense magazine.

“On National Security” appears in every issue of SpaceNews magazine. This column ran in the January 2023 issue.