A newly launched AI company has built a large language model that's intended to help the U.S. military analyze propaganda and misinformation abroad, as well as develop strategies to counteract it.
The company, called Andromeda, spun out of Austin applied sciences firm Valkyrie on May 29 with $4.5 million in pre-seed funding from a group of investors that includes Austin-based Trust Ventures, as well as several unnamed people with strong ties to the defense community.
The new money will help fuel product development and allow the startup to hire new talent for its engineering team.
The roots of Andromeda were planted after Valkyrie CEO Charlie Burgoyne talked with military leaders when the U.S. Army Futures Command was launching in Austin several years ago. That led him to embed with the U.S. military overseas and see firsthand how propaganda and disinformation played out in places such as Thailand, a longtime U.S. ally where perceptions of Americans have been on the decline.
He saw the opportunity to pair large language models, which were still mostly unknown outside of the deep tech world, and knowledge graphs. He wrote a white paper on the concept and gave it to Gen. Tony Thomas.
"I think with that technology, you'd be able to really transform the psychological operations or information warfare, and that's really the field where they synthesize what's going on with misinformation, disinformation and propaganda from foreign adversaries," he said. "I had no intention of building out the tool, but they came back and he said, 'Hey, can you guys actually start building this thing you wrote about?' And that was the first step in Andromeda."
That was 2021, and Valkyrie continued to work on government missions through 2023, when it began to conceptualize commercial applications in the oil and gas industry, as well as fintech.
Since then, AI has grown astronomically.
What sets Andromeda apart, Burgoyne said, is that its platform pairs large language models that drive generative AI applications with knowledge graphs that help it eliminate hallucinations often seen in the text and images that generative AIs produce. That's a critical step, given that its platform is used by the intelligence community to analyze misinformation campaigns and decide next steps.
Another major differentiator is that Andromeda's outputs come with citations so that military or business leaders can see the original sources that the AI used in its reasoning.
“Decision-makers in the defense community are faced with complex, high-stakes decisions on a daily basis. Filtering through piles of information sources can cost critical time, resources, and even lives” Salen Churi, general partner at Trust Ventures, stated. "We are excited to support Andromeda's work to empower these decision makers with fast-moving, fact-based insights to ensure successful high-stakes operations.”
Burgoyne sees Andromeda as a small part of a much bigger issue — how AI could dramatically alter humans' sense of self and their place in the world. It's something he spoke about during a SXSW presentation in more detail. He said the potential for AI to shift humans' relationship with the world has gotten even more pronounced since that presentation.
"I think this is the most critical conflict of the century," he said. "I think it's really, really difficult for people to be able to sift through the noise to get back to signal, and that signal is integral for their identity ... and having a relationship to their communities, their family, to their nation, to their faith. These are all things that are wildly at risk, and we're seeing really good evidence of that."
