Welcome to SpyWeek, our new weekly newsletter, where we look at news from the intersection of intelligence, foreign policy, and military operations.
Artificial Intelligence, Part I: We’ve known for some time that Artificial Intelligence is changing the U.S. intelligence business. “AI will help intelligence professionals find needles in haystacks, connect the dots, and disrupt dangerous plots by discerning trends and discovering previously hidden or masked indications and warnings,” states a 2021 report by a Congressional commission on AI. There’s been a lot of hype (and a lot of superlatives) but little explanation of how things are changing. What does life look like for a CIA officer using AI? A Day in the Life of an AI-Augmented Analyst by 32-year CIA veteran William “Chip” Usher and Tara McLaughlin tries to answer that question.
The CIA is now using artificial intelligence to help predict when critical situations of social instability are likely to occur (ISTOCK)
Usher and McLaughlin are on the staff of the Special Competitive Studies Project, a not-for-profit founded by former Google Chairman and CEO Eric Schmidt, who led the previously mentioned, Congressionally-mandated National Security Commission on AI. Usher and McLaughlin conjure up a CIA terror finance analyst tracking Hezbollah, who, for our purposes, we’ll call Jen. (Pro tip: Name your characters.) Before AI, Jen had to sort through nearly 800 pieces of intelligence daily, one at a time. Now, she can rely on an AI system that Usher and McLaughlin call “ALICE,” which stands for Augmented LLM and Intelligence Cataloging Enterprise.
ALICE would be a technological leap for the CIA, which has been exploring the uses of AI since at least 1983. The intelligence community has been using older forms of AI, such as machine learning and natural language processing tools, to help manage the flood of data in the Internet age. ALICE is the next generation of “generative AI,” algorithms (like OpenAI’s ChatGPT) that are designed to mimic how the human brain works. The algorithms find patterns in existing datasets and use those patterns to generate content such as text, images, or audio that feels increasingly real. Something like ALICE is already in the works at the CIA or may even be up and running. The CIA has revealed that it’s developing its own version of ChatGPT and is currently hiring AI specialists.
In Usher and McLaughlin’s telling, before Jen arrives for work at the CIA, ALICE has sorted the nearly 800 pieces of intelligence by color—green for low priority, yellow for medium, and red for high. Most of the intelligence is marked green, and a few pieces of HUMINT are red, but it’s a group of yellow-tagged documents that catches Jen’s eye. The yellow items—financial records, travel documents, and news reports—center around a French-Lebanese businessman in Paris. ALICE notices an unusual pattern in the businessman’s purchase of a vacation property in Morocco “that would have been difficult for a human analyst to spot quickly,” Usher and McLaughlin write. Perhaps those patterns would never have been noticed at all, but ALICE helps Jen uncover a Hezbollah financing network.
OK, so this sounds a bit like a Tom Clancy plot, but AI’s ability to find hidden patterns holds the promise of transforming intelligence.