TurbineOne, an American company specializing in intelligence analytics software, has secured a multi-year contract from the Department of Defense's Defense Innovation Unit (DIU) to create prototypes for novel machine learning (ML) capabilities in collaboration with the Office of the Under Secretary of Defense's Research and Engineering's (OUSD R&E's) FutureG and 5G Office. These capabilities will span a wide range of applications, including enhancing force protection, refining the geolocation of industrial equipment, and transforming intelligence operations centers.

In conjunction with this new contract, TurbineOne is introducing an innovative feature within its Frontline Perception System, known as "Sidekick." This functionality is designed to automatically generate data to construct new detection models. The technology employs a computer science concept called "Zero Shot," which essentially allows a computer to accurately identify objects it has never encountered before. This is analogous to the image search function on Google or video searches on Facebook, but TurbineOne's approach operates without the need for extensive web infrastructure, standardized metadata, or pre-labeled training data. Instead, TurbineOne's software will produce effective zero-shot computer vision ML models using unstructured and unlabeled data it hasn't previously encountered.

Expanding on prior collaborations with the military in the Pacific Area of Operations, TurbineOne is revolutionizing the assimilation of images from video cameras and satellite imagery, as well as non-visual data feeds from sensors. This approach, known as "multi-domain operations" and "sensor fusion" within the Department of Defense, aligns seamlessly with TurbineOne's Zero Shot detection capability's "multi-modality." While conventional computer vision systems are usually limited to specific types of training data (e.g., labeling still images to identify similarities in other still images), TurbineOne enables users to manage diverse data types from various sensor providers to construct new models within a unified, do-it-yourself platform.

Retired Lieutenant General Bob Ashley, former Director of the Defense Intelligence Agency, comments, "Having been exposed to the nation's top intelligence capabilities, I've never encountered an intelligence tool like TurbineOne's. It has the ability to find crucial information within vast datasets, significantly enhancing the effectiveness of our intelligence analysts."

TurbineOne's Sidekick feature, from a technological perspective, draws indirect inspiration from Generative AI and large language models (LLMs). However, TurbineOne relies on a precursor in data science known as ML Transformers to generate object detection models conditioned by text. This groundbreaking search experience translates queries instantly into visual objects, even within isolated data environments such as those commonly found in secure DoD networks. From a user standpoint, an intelligence analyst could simply point to a data folder and input a query like, "Display all aerial views of ships," and receive highly accurate results.

Ian Kalin, CEO of TurbineOne, remarks, "The AI-driven perception sector is still in its nascent stages, and we're honored to collaborate with DIU to advance these capabilities. In the immediate future, these tools will facilitate straightforward queries such as 'Show me all Chinese ships near Taiwan,' providing rapid and precise outcomes. In the long run, analysts will be able to pose more intricate, predictive questions like 'Which indicators strongly suggest that China has initiated the initial phase of a potential invasion of Taiwan?'"

The initial focus of the DIU partnership with OUSD R&E's FutureG and 5G Office, along with TurbineOne, will be Naval Air Station Whidbey Island. Kurt Andrews, a Principal Investigator for OUSD R&E's FutureG and 5G Office, expresses enthusiasm about the collaboration: "We're excited to collaborate with TurbineOne to develop new ML-driven geolocation and threat detection technologies. Their software platform's versatility makes it exceptionally well-suited for intelligence, surveillance, and reconnaissance applications, as it empowers users with a do-it-yourself toolkit tailored for military environments."

Source:businesswire.com