The CIA has more than 100 ongoing pilot projects involving the use of artificial intelligence (AI) but isn’t ready to adopt the technology to help with decision-making applications until it is better understands why the technology comes up with the outputs it does, according to an agency official.
“And so, one of the things that I think is a challenge for the current AI community and I’m confident we’ll get addressed, but still needs to be done, is you can’t just put things into the black box and something comes out, Dawn Meyerriecks, deputy director for Science and Technology at the CIA. “We can’t go to leadership and make a recommendation based on something that nobody understands what happened in the middle.”
Meyerriecks said that when the president’s daily intelligence briefing is put together, the evidence has to be linked to the conclusions. Speaking as part of a panel on Wednesday at the Intelligence & National Security Summit co-hosted by AFCEA and INSA, Meyerriecks said a few weeks back she discussed this challenge with some venture capitalists.
The CIA by current count has 137 AI pilots and hosted its first “community of interest” around the technology and had 100 people attend during their lunch hours from throughout the agency so it is an “Important topic for us,” she said.
Artificial intelligence refers to the ability of computers to do intelligent tasks that humans do such as decision-making, and visual and speech recognition, and includes capabilities like machine learning computers learning by accessing and using data.
Meyerriecks said she’s “confident” the AI community will be able to meet the CIA’s requirements because other markets also need the same transparency from the technology, including healthcare and legal communities. She also said that she has a “punch list” for the AI developers.
“If I want to increase my confidence, what goes into what I need in order to make a really good assessment both on the back end, because that tells me what kind of collection I need in order to raise confidence to be able to go advise national leadership,” she said.
The CIA works with organizations like In-Q-Tel, which is backed by the intelligence community, on “helping influence the industry in a way that I think benefits all of the commercial market and not just the USG,” Meyerriecks said, referring to the U.S. government. In-Q-Tel helps bring the government, technology start-ups and venture capitalists together to help meet the technology needs of government agencies.
The use of AI is in its infancy, Jeff Herbst, vice president of Business Development for NVIDIA [NVDA], one of the leading companies involved in the technology, said during the panel discussion. The technology has been applied to “low-hanging fruit” such as visual recognition and now just starting to be applied to data analytics, he said.