“We still have to validate the sources,” says Lowdon. But unit commanders encouraged the use of large language models, he says: “Because they offer much more efficiency in a dynamic situation.”
The generative AI tools they used were built by the defense technology company Vannevar Labs, which, in November, obtained a production contract Takes up to $ 99 million by the defense innovation unit focused on Pentagon startups in order to bring its intelligence technology to more than military units. The company, founded in 2019 by veterans from the CIA and the American intelligence community, joins the tastes of palanting, AndundilAnd on the scale of AI as a major beneficiary of the embrace of artificial intelligence by the American army – not only for physical technologies such as drones and autonomous vehicles, but also for software that revolutionizes the way the Pentagon collects, manages and interprets data for war and surveillance.
Although the American army has developed computer vision models and similar AI tools, such as those used in Maven projectSince 2017, the use of a generative AI – Tools which can initiate a human conversation like those built by Vannevar Labs – represent a more recent border.
The company applies existing large language models, including some of Openai and Microsoft, and some tailor -made to its own open source of the company in the company since 2021. The scale at which these data are collected are difficult to understand (and a large part of what establishes Vannevar products every day): Terabbytes of data in 80 different languages are hooverts every day in 180 countries. The company says that it is able to analyze social media profiles and violate firewalls in countries like China to obtain difficult to access information; It also uses non -classified data that is difficult to connect (collected by human agents in the field), as well as physical sensor reports that secretly monitor radio waves to detect illegal shipping activities.
Vannevar then builds AI models to translate information, detect threats and analyze political feeling, with the results delivered via a chatbot interface which is not different from chatgpt. The objective is to provide customers critical information on subjects as varied as the fentanyl international supply chains and China’s efforts to secure rare earth minerals in the Philippines.
“Our real objective as a business,” explains Scott Philips, Director of Technology at Vannevar Labs, is to “collect data, give meaning to this data and help the United States make good decisions”.
This approach is particularly attractive for the American intelligence apparatus, because for years, the world has been flooded with more data than human analysts can possibly interpret – a problem that contributed to the Palantir Foundation in 2003, a business with a market value of more than 200 billion dollars and known for its powerful and controversial tools, including a database that helps immigration and Search and follow information on undocumented immigrants.
In 2019, Vannevar saw an opportunity to use major language models, which were then new on the scene, as a new solution to the data enigma. Technology could allow AI not only to collect data, but to really talk about an analysis with someone interactively.