A trio of new investments in Silicon Valley machine-learning startups shows that the U.S.intelligence community is deeply interested in artificial intelligence. But China is investing even more in these kinds of U.S. companies, and that has experts and intelligence officials worried.
Founded to foster new technology for spies, the 17-year-old In-Q-Tel has also helped boost commercial products. (Its investment in a little company called Keyhole helped produce Google Maps.) Compared to a venture capitalist firm whose early-stage investments are intended to make some money and get out, the nonprofit’s angle is longer term, less venture, more strategic, according to Charlie Greenbacker, In-Q-Tel’s technical product leader in artificial intelligence, machine learning, natural language processing, analytics, and data science.
“Our model is to put a little bit of pressure at the right spot to influence a company to make sure it develops things that are useful to our customers,” said Greenbacker, who estimated that their investments in a given startup generally amount to about one of every 15 dollars the company has.
Greenbacker recently discussed In-Q-Tel investments that underline the specific areas that the CIA and other spy agencies want to use AI for: image recognition, natural language processing, and predictive analytics.
The first is Orbital Insight, a company that analyzes satellite images to discover trends and patterns on a global scale. As Orbital’s CEO James Crawford said last year, “The number of ships docked at a Malaysian port, even the color of a wheat field in western Nebraska, are actually signs…visible indicators of economic activity, not just for a local region but for an entire global industry.”
Natural language processing is another a big area of AI investment. Greenbacker highlighted a company called Primer, which sells a tool that can read and summarize text, useful for the CIA’s Directorate of Intelligence (now Directorate of Analysis) and the decidedly unsexy work of combing through the world’s court cases, patent filings, and newspapers for relevant data. Primer is the work of Sean Gourley, a co-creator of the Quid analysis and visualization engine and one of the more celebrated young data scientists working at the intersection of Silicon Valley and national security.
What does a spy agency do with all of this data and analysis? Try to predict the future, of course, and the intelligence community has a big interest in AI that will help it succeed there as well. In-Q-Tel invested a company called Celect, which says its predictive analytics engine can help retailers anticipate demand for certain products — and perhaps help intel analysts anticipate national-security problems. When the researchers fed the engine a large database of published news stories, it predicted, with high confidence, a steady increase in negative news about the security situation in Ukraine in the early spring of 2014. And indeed, that happened. (The experiment was done retroactively, after the Russian invasion, when outcomes were known to the researchers — but not to the AI.)
Unlike other technological areas, today’s AI startups don’t necessarily need intelligence-community money, says Greenbacker: “There’s a lot more money flowing into the AI space than is typical” in the startup ecosystem.
“But putting in a little leverage in the right spot can bend all this toward the things we care about,” he said. “At the end of the day, it means the government doesn’t have to foot all of the bill to develop technology it cares about.”
Competing with China
The intelligence community’s investment in the valley’s AI startups lags foreign investors and China. And that has some folks very concerned.
“The entire government spent $1.1 billion on unclassified AI programs in 2015. The estimate for 2016 was $1.2 billion,” said Greenbacker. “Meanwhile, Softbank [a Japanese multinational conglomerate] has a $100-billion-dollar fund for this. The Chinese government in their most recent five-year-plan has put $150 billion in this.”