
Transforming defense: South Korea develops AI-driven autonomy with synthetic data
On December 13, at the CES 2025 Global Media Meetup event hosted by AVING News MIK Basecamp in Seoul, GenGen AI, an emerging startup specializing in artificial intelligence, showcased its innovative approach to data generation tailored for the defense and mobility industries.
The event highlighted the growing significance of relevant data for deploying AI technologies, especially in sectors where data security is paramount.
CEO Cho Ho-jin of GenGen AI introduced the concept of synthetic data—artificially created data that stands in for real-world information.
He emphasized that while the autonomous driving industry has made significant progress utilizing synthetic data, the defense sector is just beginning to explore these options.
AI-driven autonomy
“Autonomous driving has been a pioneer in synthetic data application due to the extensive collection of actual road data in places like Korea,” Cho noted. “However, the challenges that arise in data collection can be mirrored in the defense sector.”
Cho explained that in autonomous driving, vehicles are equipped with cameras to gather real-time data. However, variations in legal regulations, road signs, and potential accident scenarios across different regions complicate matters. Even industry leaders like Tesla face difficulties; their vehicles sometimes struggle to recognize animals on the road, leading to accidents.
GenGen AI creates diverse synthetic images to address these challenges, ranging from clear weather scenarios to various environmental conditions like rain and snow, alongside additional animal data.
This synthetic imagery is then supplied to automotive manufacturers and parts suppliers.
For a notable project in China, GenGen AI generated realistic Chinese traffic signs tailored to multiple scenarios, achieving an impressive accuracy rate of around 99%.
Looking at the future, Cho mentioned that market analysts expect synthetic data to multiply, potentially surpassing reliance on real data.
