[On the Spot] “Technology Is Ready” — Funzin Emphasizes Practical Application of Defense AI
Funzin unveiled practical applications of Artificial Intelligence Internet of Things (AIoT) technology in the defense sector, highlighting the need for institutional and operational reviews to promote the technology’s adoption in the future.
As the AI-based decision-making system has reached a high level of technical maturity, it is now time to begin discussions on its practical implementation in earnest.
At the 51st Korea Artificial Intelligence Industry Association (AIIA) Breakfast Forum, held at L Tower in Yangjae, Seoul on Thursday, Funzin CEO Kim Deuk-hwa announced the company’s defense AIoT technology strategy.
According to Kim, Funzin’s Kill-Web Matching (KWM) system was introduced on a trial basis in the Army’s Tiger Unit, verifying a real operational case that demonstrated a rapid process—from AI-based target recognition to weapons recommendation and command transmission.
The core concept is an “AI Staff”, which integrates sensors and data flows on the battlefield to support faster command decision-making.
The core technology is object recognition based on few-shot learning, which utilizes small amounts of data.
Funzin developed a lightweight model capable of operating without large training datasets, enabling real-time detection and classification even in embedded environments.
The company also emphasized the potential for effective data augmentation within closed networks through its proprietary synthetic data generation platform “Eagle Eye,” which has recently completed an actual suitability assessment.
The AI command system consists of four main processes: target detection, weapon recommendation, command approval, and command transmission.
Funzin integrated data from 13 types of equipment, including drones and armored vehicles, and reduced the complex command procedure—from more than 90 minutes during initial demonstrations—to under 10 minutes.
CEO Kim emphasized, “However, AI does not make direct judgments. The focus was on assisting human decision-making and refining available options.”
It was further highlighted that the company is technically compensating for data scarcity.
Because it is difficult to secure actual equipment data in the defense environment, few-shot learning and synthetic data–based learning structures are being applied.
A technical roadmap has also been proposed to gradually improve model accuracy through accumulated data.
The company also mentioned the potential to expand into non-image-based detection systems, such as electronic warfare and radar.
Funzin is developing a learning model optimized for the battlefield environment, introducing physics-based AI technology, which has recently drawn attention.
Unlike generative AI, this approach focuses on enhancing real-world applicability by incorporating physical characteristics.
“In the defense sector, AI should function as a support system that helps organize judgments and decisions more quickly,” said Kim Deuk-hwa, CEO of Funzin.
“As lightweight and on-device technologies suitable for field operations become available, we expect cooperative discussions for practical deployment to become more active.”
댓글
댓글 쓰기