Krafton unveils Ludi 0.2 robot AI to interpret human intent

A leap in social intelligence

Ludo Robotics, a subsidiary of the South Korean company Krafton, recently debuted its Ludi 0.2 robot AI system. The unveiling took place at the 2026 Bay Area Robotics Symposium at UC Berkeley. This new system is designed to handle the complexities of physical AI, which connects generative AI to actual robot control in the real world.

Unlike its predecessor, Ludi 0.1, the new model focuses on higher-order social intelligence. While the first version combined vision-language models for reasoning with basic movement, Ludi 0.2 is built to understand human gestures and the context of a conversation. This allows the robot to behave more naturally when interacting with people in shared spaces.

The system is specifically engineered to respond to unexpected situations and feedback. It can process social cues to determine what a human wants even when instructions are not perfectly clear. By improving memory and reasoning functions, the AI maintains the context of a situation even as the environment or the user's demands shift.

Adapting to changing requests

One of the primary goals of Ludi 0.2 is its ability to handle mid-task changes. In a practical example, if a robot is already moving to fetch a specific drink and the user suddenly asks for a different one, the robot can stop its current action. It then re-assesses the request, selects the correct beverage, and completes the new task.

This flexibility is a significant step forward for robotics because it moves away from linear, pre-programmed paths. The AI must constantly evaluate the human's intent rather than just following a single command to completion. This capability is essential for robots intended to work in homes or offices where human needs are fluid and often change on the fly.

The technical foundation of this behavior relies on the robot's ability to integrate vision and language. By understanding the nuances of how people speak and move, the AI can bridge the gap between a spoken command and a physical action. This ensures that the robot remains useful even when the user provides feedback or corrects a mistake during a process.

From games to physical AI

Krafton is leveraging its deep history in the gaming industry to build these robotic capabilities. The company has spent decades developing real-time interactions within virtual worlds. They are now applying that knowledge to help robots communicate and cooperate with people in physical spaces.

This transition marks a strategic move for Krafton, which has traditionally focused on entertainment. By entering the physical AI space, the company aims to create a new growth engine for its business. They are using their expertise in simulation to solve the difficult problems of real-time interaction and movement control.

To support this expansion, Ludo Robotics has established a new research headquarters in Palo Alto, California. This location is intended to facilitate deeper exchanges with the talent and technology ecosystem in Silicon Valley. The company will continue to conduct its primary AI and robotics research between this new hub and its existing office in Seoul.

The path toward Ludi 1.0

While Ludi 0.2 is a significant upgrade, it is still considered a research-stage system. Krafton has not yet disclosed a specific timeline for commercial use or a revenue model for the technology. The current focus remains on refining the AI's ability to process complex human intent and smooth physical movements.

The long-term goal for the company is the development of Ludi 1.0. This next-generation foundation model is intended to be a unified system that processes all robotic functions through a single neural network. This would mean that cognition, dialogue, and physical control would all happen within one integrated framework rather than separate modules.

The robotics market is currently highly competitive, with major semiconductor and platform companies like Nvidia also vying for dominance. Krafton faces the challenge of securing a technological edge in this crowded field. The success of their strategy will depend on whether they can successfully connect their research to a viable product ecosystem for consumers.