Summary of Key Points
The second World Humanoid Robot Competition was held in Beijing, with over 600 teams participating. It showcased the “superhuman” capabilities of robots in athletic events such as sprinting, high jumping, and backflips (for example, a robot ran 100 meters in 8.64 seconds, faster than Usain Bolt, and jumped 3.4 meters, breaking the human record). However, it also revealed their “clumsiness” in everyday tasks like driving a nail or folding clothes. This competition is not just about showing off technical skills but also a test of the robots’ “practicality.” To integrate into human life, robots must overcome three major challenges: dexterity, safety, and cost. Physical AI models and simulation training are accelerating their development.
1. Robots’ Olympic Performance: Dominating in Athletics with Pre-programmed Movements
The robots’ performance in athletics was astonishing: a robot from the Beijing Humanoid Robot Innovation Center ran 100 meters in 8.64 seconds, nearly a second faster than Bolt’s world record; another robot jumped 3.4 meters, breaking a record that had stood for over 30 years. How can robots outperform humans in these tasks? Stanford scientist Karen Liu explains that these movements are based on “pre-programmed trajectories”—like a precisely choreographed dance, with every detail (what to do and how much force to use) calculated to the thousandth of a second, eliminating the need for “thinking.” This fixed approach allows robots to excel in athletics but limits their ability to handle complex tasks.
2. Daily Tasks Challenge the “Powerhouses”: Simple Tasks Prove Difficult
In contrast to their athletic achievements, robots struggle with everyday tasks like driving a nail or folding clothes. For instance, some robots struggle to even drive a nail, and even with remote control, they cannot perform the task accurately. The reasons lie in two main issues: first, their “hands” lack the flexibility to control force and angle accurately; second, their “brains” are not capable of real-time adaptation to the environment (e.g., handling clothing wrinkles or nail positions). Humans can adjust their movements based on the task, but robots cannot.
3. Technology Acceleration: Physical AI Models + Simulation Training
To teach robots practical skills, scientists use “physical AI models.” These models convert human movement data into commands that robots can understand. For example, by having humans wear VR helmets and sensors while folding clothes, robots can learn the process. They can also practice in simulated environments before moving to the real world, reducing errors and costs. Boston Dynamics’ Durbin notes that this approach is rapidly improving robot performance, such as in their Atlas robot, designed for tasks humans dislike (like folding clothes or working in hazardous environments). Beijing Acceleration Technology is also developing robots for repetitive, physically demanding tasks.
4. Three Barriers to Integration into Daily Life
For robots to become useful, three major hurdles must be overcome:
1. Safety: Robots are often heavy (some weigh several kilograms), and a system failure could result in injuries. Carnegie Mellon researcher Atkeson dreams of creating lighter, more flexible robots that won’t harm humans.
2. Cost: The cheapest humanoid robots cost around $5,000 (about 35,000 RMB), and more advanced models are even more expensive, making them unaffordable for most people.
3. Reliability: Robots often make mistakes, such as crashing into walls, tripping, or starting fires, indicating a lack of stability.
5. The Competition’s Role in Transitioning from “Toys” to “Tools”
This competition is more than just a showcase of technology; it’s a catalyst for making robots more practical. Tasks like organizing bookshelves and making beds were included, with robots that complete these tasks independently receiving higher scores. This encourages researchers to focus on real-world applications rather than purely athletic challenges. Durbin emphasizes that robots must solve practical problems, such as helping with household chores or dangerous work.
In summary, humanoid robots have impressive athletic abilities but struggle with practical tasks. With ongoing technological advancements and a focus on practical applications, they could become valuable helpers in the future. However, they need to improve their dexterity, safety, and cost-effectiveness before becoming a common part of our lives.