The Central Intelligence Agency (CIA) wants to mature its multi-cloud environment for advanced artificial intelligence initiatives, but it needs industry partnerships to do so, a top Federal official said on Sept. 5 at the Billington CyberSecurity Summit in Washington.
The CIA will increasingly explore the use of generative AI to streamline its operations in human-machine teaming, according to Juliane Gallina, the deputy director for digital innovation at the CIA.
“We need all variations of the cloud, everything from hybrid cloud, private cloud … we still need high-performance compute,” said Gallina. “We may also still continue to need on-premise compute for all of these things, for the full spectrum of our performance.”
“We need multi-cloud to really become fully mature with generative AI to use large language models or diffusion models to really optimize our performance,” she continued.
The growing interest in generative AI is because it is far more “intuitive for end users” compared to traditional machine learning approaches, Gallina explained.
This has signaled a shift in how the intelligence community approaches data and human-machine learning – especially with the mass amounts of open-source data. Gallina hinted that generative AI will play a larger role in CIA operations in the future.
“[Generative AI is] much more viral than traditional machine learning approaches were for us, so I think in the future, we’re going to continue to have both in our enterprise,” said Gallina.
The CIA was the “pioneer” of cloud usage in the Federal government, first using multi-cloud environments in the early 2010s, Gallina said. She explained that tech innovation and usage is in the agency’s “DNA,” and that they want to continue to innovate.
She called out to potential industry partners, saying, “If you’re in industry and you’re listening, you’re eager to learn, and you’re curious about us, we’re curious about you too.”
![Weslan Hansen](https://www.meritalk.com/wp-content/uploads/2024/07/Weslan_Hansen_Headshot-e1721246033890.jpg)