Rapidly Growing Humanoid Robot Industry with AI

Tech companies are fiercely competing for technology to make humanoid robots more human.

Until now, robots used in the field have been modeled after human arms or in the form of machines that move heavy objects.

Some humanoid robots were introduced, but they were only at the level with show-off functions such as climbing stairs and spinning in the air.

However, as it has become possible to learn natural behavior by collecting large-scale human behavior data using artificial intelligence (AI) and interacting with people or the environment around them, the development of humanoids is rapidly accelerating.

As the number of humanoid robots used in actual industrial sites increases significantly, the humanoid robot market, which was worth 1.8 billion dollars (about 2.4 trillion won) last year, is expected to grow rapidly to 13.8 billion dollars (about 18.4 trillion won) by 2028, according to market research firm Markets & Markets.

Just as the key factor influencing the performance of AI models is the amount of training data, how much humanoid robots also learn human behavior data is important.

Tesla recently put out a job posting to recruit "data collection operators." Their primary task is to wear motion capture suits and VR devices for more than seven hours a day and perform designated tasks.
In order to collect effective behavior data, the condition was taken that the height should be 170-180 cm, similar to Optimus.

The hourly wage is 25.25 to 48 U.S. dollars (about 30,000 to 60,000 won). Business Insider, an economic media outlet, said, “Tesla aims to collect vast amounts of data to train robots.”

Experts estimate that hundreds of thousands to millions of hours of data must be learned in order to put robots into factories. Dozens of data collection personnel are hired separately, just like donating blood, because there is not enough data to get close to people.

When Tesla unveiled Optimus in December last year, it only walked and held eggs. However, when it was unveiled in May, it became possible for Optimus to classify products while looking at them.

Some predict that if we learn the movement of human fingers as it is currently in progress, we will be able to participate in the actual car assembly process. Recently, Musk announced that he will put Optimus into the Tesla factory next year and sell it in 2026.

In recent humanoid development, robot companies are paying the most attention to 'hand' and 'brain'. Aptronic, an American robot startup, has developed a humanoid 'Apollo' that uses only electrical components to control five fingers rather than using pressure.

Electrical components are small and can be applied to every knuckle of a robot's hand. As such, sophisticated control is possible, complex tasks such as machine assembly can be performed.

AI plays the role of brain of humanoid robots. Figure AI, a U.S.-based robot startup, uses Nvidia's graphics processing unit (GPU) to triple the inference performance of AI on humanoid robots.

Robots have the ability to judge and process in real-time in actual work. Recently, the company's robot "Figure 02" has been deployed in BMW's car production line.

Developed by Shankturi AI, 'Pinix' has significantly enhanced the visual and tactile functions of robots. Thanks to this, it can provide high-quality data to AI.

Learning new tasks has reduced the time taken from weeks to less than a day. Robot company 1X is developing a humanoid robot that combines OpenAI's Generative AI ChatGPT.

Goldman Sachs said, "The main task is to develop AI for interaction, such as taking orders from people without training."

Key Takeaways: Humanoid Robots and AI

• Rapid Growth: The humanoid robot market is rapidly expanding, driven by advancements in AI and the ability to collect and learn from massive datasets of human behavior.
• Learning from Humans: Companies are actively investing in data collection efforts, using motion capture suits and VR to capture human movements for robot training.
• Focus on Hands and Brains: Developing more sophisticated robotic hands and brains is crucial for advanced tasks, leading to advancements in hand control, AI processing power, and visual/tactile perception.
• Applications in Manufacturing: Humanoid robots are increasingly being deployed in industrial settings, with potential applications in car assembly and other tasks.
• Future of Humanoid Robots: Companies are aiming for robots that can interact naturally with humans, take orders, and learn new tasks quickly, bringing us closer to a future where robots are integrated into daily life.
• Importance of AI: AI plays a central role in enabling real-time judgment, processing, and task learning in humanoid robots.