Silicon Valley "All In" humanoid robot
Silicon Valley tech giants such as Amazon, Tesla, Microsoft, Meta, and Apple are actively developing humanoid robots, despite facing challenges such as data training, safety, and reliability. Humanoid robots require a large amount of training data to achieve human-like interaction, with main training methods including remote operation imitation, simulation training, and hybrid training. Safety issues cannot be overlooked, especially in home environments. The high cost and technological limitations of humanoid robots pose difficulties for their widespread adoption, which may take decades to achieve
According to media reports on Monday, humanoid robots have recently become a new battleground for tech giants, with Amazon, Tesla, Microsoft, Meta, and Apple all investing heavily in the development of humanoid robots.
However, there are still many challenges to overcome to realize this vision, among which massive data training, safety, and reliability are two significant hurdles. The widespread adoption of humanoid robots may take decades.
First, in terms of training data, the biggest feature of humanoid robots compared to existing robots is their human-like interaction methods. This is reflected not only in how they interact with objects but also in how they recognize objects and understand how to use them, which is the concept of "affordance" in the field of robotics.
When a robot sees a cup, it needs to know that it can grab the handle and then drink or pour water. This requires combining the robot's hand, computer vision, and reasoning abilities.
To enable robots to understand how to interact with objects, a large amount of training is needed. Currently, there are three main training methods:
Remote operation imitation: Humans wear special equipment to perform tasks, and the equipment collects human motion data and transmits it to the robot, guiding the robot on how to move.
Simulation training: Jensen Huang, CEO of Nvidia, demonstrated a simulation training method called Nvidia Cosmos during his keynote speech at CES 2025. Professor Agrawal from MIT's Department of Electrical Engineering and Computer Science pointed out that this method can create various scenarios in the digital world and translate knowledge into robot behavior. Its advantages are speed and safety, but the downside is that there is a gap between the simulated environment and the real world, and it is difficult to simulate soft, deformable objects like clothing.
Hybrid training: This combines real-world human actions with simulation training, but this method is still under research. Researchers are trying to merge the advantages of both methods.
In addition to training, the safety and reliability of humanoid robots are also critical issues. Ye Zhao, an assistant professor in the School of Mechanical Engineering at Georgia Tech, pointed out that if a humanoid robot falls at home, not only are the repair costs high, but it can also endanger the safety of people in the home.
Currently, the cost of humanoid robots can reach tens of thousands of dollars, and they have short battery life and are prone to overheating. Although some companies claim to launch lower-cost humanoid robots, the actual costs remain high, and issues such as overheating and unreliability persist.
Risk Warning and Disclaimer
The market has risks, and investments should be made cautiously. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk