[Original Article] "생성형AI 다음은 피지컬AI… 머리 빗고 명함 집는 섬세한 로봇팔 만들것"

"The next most promising area after generative AI is physical AI. We are currently focusing our capabilities on this field." In an interview with our newspaper at the company's headquarters in Magok, Gangseo-gu, Seoul, early this month, Kim Byung-soo (pictured), CEO of ROBOTIS, said, "Physical AI is a function in which robots will perform functions that humans used to do with their arms." ROBOTIS is a company specializing in autonomous robots, of which LG Electronics is the second largest shareholder.

Physical AI refers to robots that apply AI to robotic arms (manipulators), such as so-called "sensory moving arms." CEO Kim gave the example of combing hair. You need to press the comb with an "appropriate" amount of force, but current robots only move according to established logic, making "appropriate" impossible. Physical AI means that this "appropriate" thing is possible.

ROBOTIS' robot actuator 'DYNAMIXEL' on display at ROBOTIS' headquarters in Magok, Seoul. Reporter Jang Woo-jin

Another example is picking up a business card. In the current method, a robot arm has to find the location of the business card, move it precisely, adjust the angle, and pick it up with a fingernail. However, people do not pick up business cards in this way. It is difficult to say that the brain picks it up, rather than the hand.

CEO Kim said, "Previously, robot arms operated based on location, so they were unable to do things that were fixed, such as opening doors. Most daily tasks, such as changing pillow sheets, brushing teeth, and combing hair, are fixed." He added, "If robots can perform functions that humans used to do with their arms, we will be able to reexamine the value of AI, thinking, 'AI really benefits us. '

" He continued, "The ultimate answer for robots is that they must move based on physical AI, but the problem is how robot arms should be made to do that." He explained, "Just as muscles are needed depending on the purpose, robots are no different. In daily life, they do not need to be as strong or fast as general industrial robots, and they should not be."

He emphasized, "In terms of power consumption and safety, the robot arm should be equipped with sensors, such as cameras, that can observe its own role, appropriate for the purpose." He added, "With that, the main board controller needs to be able to move the arm on its own without having to individually control its position."

CEO Kim explained that physical AI began to emerge from the Aloha Project of Stanford University and Google DeepMind. Stanford University released the 'Aloha' robot developed based on 'DYNAMIXEL' on YouTube in January of this year, and it became a hot topic in the related industry.

A screenshot of the Aloha Project by Stanford University and Google DeepMind. The robot arm is naturally inserting the chair. Screenshot of Stanford University YouTube video.

The Aloha robot in the video has two robotic arms attached to a wheeled shelf. It appropriately oils and grills shrimp, and also turns them over. In the restaurant, the robot neatly arranges chairs that are outside, and gently lifts a glass filled with wine with one arm and wipes the water flowing down the glass with the other arm. Every movement is very natural, like a person.

In particular, the arm attached to the Aloha robot is ROBOTIS' 'DYNAMIXEL', and it has attracted attention from the industry along with the video. ROBOTIS is also considering the development and commercialization of a robot platform that can be directly utilized in real life like 'Aloha' by installing 'physical AI' with MIT.

CEO Kim said, "Our DYNAMIXELs are being utilized in the Aloha project, and we have felt clear signs of change in sales since the project was released." He added, "What our company is contemplating is the next-generation collaborative robot, and the basis for that is physical AI. We believe that we can play a bigger role in the hardware field, but to do that, we need to run many AI algorithms, so our competitiveness is very high."

CEO Kim said that it is difficult to predict when the physical AI era will begin in earnest, but he predicted that there will be demand from developers around 2025~2026. He predicted that from this process, when the full-fledged physical AI era opens, ROBOTIS' role with core technologies will be highlighted.

A screenshot from the Aloha Project by Stanford University and Google DeepMind. A robotic arm is holding a wine glass while the other arm is cleaning the floor. Screenshot from Stanford University YouTube video.

He said, "As Nvidia emerged as the company with the largest profit with the advent of the generative AI era, companies that supply core components will also benefit in the era of physical AI." He added, "Companies that produce core components will also stand out in the era of physical AI. Our company has a clear competitive edge because we directly produce reducers, which account for the largest portion of the production cost of robot arms."

He added, "I don't think it will be easy for humanoid robots to generate actual added value in the short term," and "They don't necessarily have to have two legs or two arms, like humans. It will be a priority to replace labor by increasing efficiency by moving on wheels or attaching three to four arms."

After graduating from the Department of Electrical Engineering at Korea University in 1993, CEO Kim stood out in the robotics industry by winning numerous robotics competitions, including the All-Japan Micro Mouse Competition and the Robot Soccer Competition, between 1995 and 1999. Since founding ROBOTIS in 1999, he has achieved major results such as reaching the $10 million export milestone and winning the Presidential Award, and was listed on KOSDAQ in 2018. He currently serves as a director of the Korea Robot Industry Association, director of the Korean Robotics Society, and president of the Steam Education Association.