Data is competitiveness, and the success of physical AI hinges on its ability to collect data.

At the seminar titled 'Physical AI, a Game-Changer Beyond Robot Limitations' held on the 11th and hosted by Safetics, Hongseop Choi, CEO of Maum AI, presented on the 'Manufacturing-Centered Physical AI Introduction Strategy' and suggested data collection as the key to the success of physical AI.

CEO Choi Hong-seop asserted, "Physical AI, which has emerged as a new turning point in the AI industry, goes beyond simply providing knowledge in the digital world to enable robots to move and perform actions in the real world. However, for Physical AI to truly achieve results, simple algorithms or model performance alone is not enough. The key is data collection. Without massive amounts of data, robots cannot learn and adapt in real-world environments, and commercialization is also impossible."

He continued, “The foundation of physical AI is the LLM (Large-Scale Language Model) and its extended VLA (Vision-Language-Action) model. “While LLM understands language and plans, VLA combines vision and action to perform real tasks, but all of this ultimately depends on the quality and quantity of data,” he said.

For robots to move in various environments, a large amount of sensor data and behavior records are required.

All defect detection, material handling, and assembly processes occurring in manufacturing sites must be accumulated as data.

Without enough data, robots will go beyond simple simulations and repeat errors in real-world environments.

In other words, a physical AI without data has enormous potential, but it is nothing more than a "blind giant" that has lost its way.

According to CEO Choi Hong-seop, the field where physical AI will have the greatest impact is manufacturing.

Large companies, such as Samsung Electronics' semiconductor factories, with automation rates reaching 98%, already have traditional automation systems in place.

On the other hand, more than half of all manufacturing still requires human intervention. What is needed at this time is not simply the introduction of robots, but a process of systematically collecting and learning data from the field.

In industries like the food and beverage (F&B) where raw materials are not uniform in form, data must be accumulated to enable robots to adapt to various situations.

Small and medium-sized manufacturers can automate repetitive tasks and leverage general-purpose robots through data collection.

Ultimately, manufacturing innovation starts with data, and the success of physical AI hinges on data collection capabilities.

Global company NVIDIA plays a key role in the physical AI ecosystem.

Their Omniverse simulator reproduces real-world data in a virtual environment, helping robots learn from a variety of situations.

The fact that domestic companies have succeeded in rapid commercialization through collaboration with the NVIDIA ecosystem is ultimately due to their effective collection and utilization of data.

Data collection and simulation are not just technical support; they are essential strategies for surviving in global competition.

Tesla's Elon Musk predicted that 10 billion humanoid robots will be in use by 2025.

This means a market worth hundreds of thousands of won.

On the other hand, simply producing robots is not enough to become a winner in this huge market.

CEO Choi Hong-seop asserted, “Only companies that can answer the questions of who secured more data, who trained in more diverse environments, and who commercialized data more quickly can take the lead in the physical AI market,” and “Data is competitiveness, and the success of physical AI depends on the ability to collect data.”

CEO Choi Hong-seop said, &ldquoFor physical AI to truly demonstrate its value, collaboration between humans and robots is necessary. For robots to collaborate with humans on production lines, data on various collaborative situations must be accumulated," he said. "The same holds true for robots collaborating with each other. Data is not simply a record; it is the foundation that enables an intelligent collaborative ecosystem. Ultimately, the future of physical AI will be determined by its ability to collect and utilize data."