제 29 호 Physical AI
Kicker: SOCIETY
Physical AI
By Ji-Won Choi, reporter
그림 1Atlas
On January 5, Hyundai Motor Group unveiled its next-generation electric humanoid robot, “Atlas,” at the CES 2026 Media Day held in Las Vegas.
Atlas was revealed in two versions: a research model and a development model. The research model demonstrated tasks by walking naturally across the stage like a human, waving to greet the audience, and rotating major joints—including the neck, shoulders, and waist—through a full 360 degrees.
The development model for industrial sites is equipped with tactile sensors and can recognize all directions through a 360-degree camera. In addition, through artificial intelligence, it can independently learn most tasks within a single day, and when the battery runs low, it moves to a charging station on its own, replaces the battery, and resumes work. This intelligent system enables the maximization of operational efficiency.
Definition of Physical AI
Atlas can be described as a form of physical AI. Physical AI refers to AI systems that operate and move in the physical world. Through sensors, they see and hear the real world, understand and analyze situations, and then control robots or machines to interact with their surroundings.
Until now, we have used AI by entering text or images to request help in certain situations. Based on that input, AI analyzed the information and provided answers. However, physical AI is not limited to the digital space. It directly observes and interprets the real world and creates physical effects through action.
Physical AI is considered the next stage of the AI industry. At CES 2025, the world’s largest AI exhibition, Jensen Huang, CEO of NVIDIA, said, “The next frontier after AI is physical AI.”
Differences from Traditional Robots
Then how is physical AI different from traditional robots? Traditional robots are machines that repeat rules and movements that have been pre-programmed by humans. When the environment changes, they may stop operating or require humans to reset them. For this reason, they are mainly used in factories and production lines.
In contrast, physical AI can recognize and evaluate complex real-world situations with many variables. The AI then controls the robot’s body. Through cameras and sensors, it observes its surroundings, calculates what needs to be done in the current situation, takes action, and even makes its own plans. For example, it can decide how strongly to grip an object or how to respond when a person suddenly walks by, making decisions based on multiple factors at the same time.
The Development Process of Physical AI
AI in industry has developed in several stages. First, it began with AI that could understand images, words, and sounds. The second stage was generative AI, which creates text, images, videos, and audio based on what it has learned and recognized. The third stage is physical AI, which can reason, make plans, and turn those plans into real-world actions.
Fields Where Physical AI Is Used
Physical AI is applied in manufacturing and smart factories, where it performs defect inspections as well as precise assembly and welding tasks. In logistics and retail environments, it transports and sorts goods while safely moving without colliding with people. In autonomous vehicles and robotaxis, sensors detect driving conditions and control the vehicle accordingly, while drones not only deliver packages but also carry out high-risk tasks such as facility inspections, surveillance, and surveying.
In the medical and healthcare field, surgical robots analyze a patient’s condition and precisely control medical instruments, while also assisting in rehabilitation and caregiving. In energy, infrastructure, and smart city systems, physical AI manages traffic and resources in an integrated way and handles maintenance tasks. In agriculture, autonomous farming machines and smart farm management systems help address labor shortages. In national defense and public safety, drones and unmanned vehicles monitor threats around the clock and are used as autonomous weapon systems to carry out operations.
Types and Forms of Physical AI
그림 2Proteus
First, transport robots move goods by finding their own paths inside warehouses or factories. A representative example is Amazon’s autonomous mobile robot, Proteus. Second, there are manipulation robots. Unlike traditional robotic arms that simply repeat fixed motions, these robots observe their environment, make judgments, and take appropriate actions. Third, mobility systems use AI to control transportation devices such as cars and other vehicles used for traffic.
Fourth, there are general-purpose robots. These robots are designed to respond appropriately in various situations, and humanoid robots are a example of this type.
Industrial Changes Caused by Physical AI
The spread of physical AI is expected to go beyond simple automation and fundamentally change modern industrial structures and working methods. Experts predict that “dark factories,” where robots and AI operate production processes without human workers, will expand, leading to the unmanned transformation of many industrial sites. From the industry’s perspective, the adoption of robots and AI is seen as a driving force that can increase productivity and strengthen national competitiveness. Increased productivity may also create new demand for goods and services, which could lead to the expansion of new job opportunities. On the other hand, labor groups express concern that these changes may restructure the labor market and deepen inequality. As seen in discussions surrounding the introduction of Hyundai’s Atlas, there are claims that new technologies could reduce the roles of existing production workers.
Physical AI presents both an opportunity to increase industrial productivity and a challenge of restructuring the labor market at the same time. As technological advancement is an inevitable trend, the redefinition of human roles, along with social consensus and institutional responses, must proceed together in order to achieve sustainable industrial transformation.
Source:
https://www.forbeskorea.co.kr/news/articleView.html?idxno=401288
https://newneek.co/@newneek/article/38750
HyundaiWorldwide Youtube
https://www.ti.com/ko-kr/about-ti/company/case-study/amazon-robotics.html
