Photo credit: www.technologyreview.com
Vannevar Labs: Transforming Military Intelligence with Generative AI
“We still need to validate the sources,” comments Lowdon, acknowledging the crucial step of ensuring credibility. The unit’s leadership has shown strong support for leveraging large language models, asserting that these AI tools significantly enhance operational efficiency in fast-paced scenarios.
The advanced AI tools in question were developed by Vannevar Labs, a defense technology firm that secured a production contract with the Pentagon’s Defense Innovation Unit last November, valued at up to $99 million. This contract aims to extend the reach of its innovative intelligence solutions across various military units. Vannevar Labs, established in 2019 by former CIA and intelligence professionals, stands alongside other prominent companies like Palantir, Anduril, and Scale AI as a leading entity in the U.S. military’s pursuit of artificial intelligence. This effort encompasses not only physical technologies such as drones and autonomous systems but also sophisticated software that is transforming how the Pentagon gathers and interprets data for military operations and surveillance.
While the U.S. military has been at the forefront of AI initiatives since 2017, notably with projects like Project Maven, the recent incorporation of generative AI, capable of human-like interactions, marks a significant evolution in this field.
Vannevar Labs employs a mix of existing large language models from OpenAI and Microsoft, alongside proprietary creations, to process vast amounts of open-source intelligence collected since 2021. The scale of data aggregation is remarkable: terabytes of information in 80 languages are gathered daily from 180 countries. The firm claims to analyze social media and penetrate barriers in nations like China to obtain information that is not readily accessible. Their methodologies include utilizing nonclassified data acquired through human operatives and reports from sensors monitoring radio waves to identify suspicious shipping activities.
The AI models developed by Vannevar focus on translating information, pinpointing potential threats, and assessing political sentiments. These insights are shared with users via a chatbot interface similar to ChatGPT, providing vital information on diverse issues such as international fentanyl trafficking routes and China’s endeavors to secure rare earth materials in the Philippines.
According to Scott Philips, Vannevar Labs’ chief technology officer, the core mission is “to collect data, make sense of that data, and help the US make good decisions.” This approach resonates deeply with the U.S. intelligence community, which has struggled to manage the overwhelming influx of data that surpasses the capacity for human analysis—a challenge that was partly responsible for the creation of Palantir in 2003. Palantir has grown to a valuation of approximately $217 billion, lauded for its powerful and sometimes controversial tools, particularly its database utilized by Immigration and Customs Enforcement to track undocumented individuals.
Recognizing the burgeoning potential of large language models in 2019, Vannevar Labs sought to harness this technology as an effective solution to the data management dilemma. The capability for AI not only to gather data but also to engage interactively in analytical discussions presents a new frontier in military intelligence.
Source
www.technologyreview.com