“Virtual game environments are ideal for robot training” After Hyundai Rotem, now partnering with POSCO DX Launching development of a robot foundation model
The NCSoft subsidiary NC AI, developer of the popular game “Lineage,” has formally entered the “physical artificial intelligence (AI)” market. The decision is based on the view that robots can be trained efficiently using reality-based simulation technology accumulated as a game company. Following earlier collaboration with Hyundai Rotem, NC AI announced plans to expand its physical AI business through a new partnership with POSCO DX.
NC AI announced on the 31st that it signed a strategic memorandum of understanding (MOU) with POSCO DX on the 29th of last month for joint development of a robot foundation model and technological cooperation. The two companies plan to pursue cooperation in areas including △joint research on robot foundation models △building a digital-twin-based simulation environment. NC AI intends to focus on optimizing an AI model that integrates vision, language, and action processing (VLA), while POSCO DX plans to construct a test environment based on digital twins.
The performance of physical AI applied to robots or factories depends on how much the gap between virtual simulations and the actual physical environment can be narrowed. In this respect, AI companies rooted in game development such as NC AI have a major advantage in that they possess sophisticated simulation environments. As game-based virtual worlds increasingly resemble reality, they provide an optimal space for effectively training robots.
Building on this strength, NC AI has developed a “world model” that teaches robots the laws of physics. The world model is a core technology for realizing next-generation robot foundation models. The company previously formed a consortium with Hyundai Rotem and, on the 28th of last month, was selected as the final contractor for the “physical AI-based integrated simulator and modular robot system” project commissioned by the Agency for Defense Development (ADD).
Choi Ji-won
AI-translated with ChatGPT. Provided as is; original Korean text prevails.
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.