China Exportsemi recently reported on the world's first humanoid robot half marathon on April 19 (related reading:From Zero to One: China’s Robot Industry Is Rising in the World’s First Half-Marathon). The event was co-sponsored by the Beijing Station of China Central Radio and Television, attracting many universities, enterprises and scientific research institutions to participate in the competition with self-developed humanoid robots. It is not only a technical competition, but also an excellent window to observe the current level of development of humanoid robots.
Surprisingly, after the competition, the runner-up "Noetix RoboticsN2" robot naughty boy was auctioned off on the Jingdong platform. The starting price was only 1 yuan, and the final transaction price was only 56,800 yuan, which was far lower than its official selling price (88,000 yuan). Although the number of onlookers exceeded 40,000, this "Taobao auction in the technology world" was particularly deserted.
This competition and auction, as well as the "technical cold thinking" behind it, unexpectedly revealed the true state of the current development of humanoid robots. Under the bustle on the surface, there is still a long way to go from the goal of entering thousands of households.
Ⅰ. Who is the test of the same hardware and different results?
The editor of China Exportsemi found that one of the most noteworthy phenomena of this marathon is: multiple teams use the same model of humanoid robot ontology - for example, there are multiple Unitree G1 humanoid robots on the field, due to the unsatisfactory results of the competition, it has also caused a lot of turmoil, and Unitree officials also came out to declare that "they did not participate"; Another example is the "Noetix Robotics" small whirlwind with the same style as the runner-up little urchin, etc. But in the race, the same "hardware" ran completely different results. This can't help but trigger the editor's thinking: the core of technology is "soft" rather than "hard". On the surface, it is compared to the machine, but in fact, it is compared to the algorithm design, sensor fusion, motion control, gait planning and other capabilities of the team behind it. The participating robot is more like a "programmable platform", and the level of the programmer directly determines the final performance of the robot.
Pictured: The first humanoid robot in the half-marathon field of Noetix RoboticsN2 robot
This also reflects that the current humanoid robot is still in the stage of high dependence on human operation and programming, and there is still a significant gap from "autonomous intelligent running".
Ⅱ If the auction is not popular, is it because consumers are "not interested"?
A robot that ran into the runner-up was auctioned off and sold for only 56,800 yuan, what problems did this expose?
1. Audience enthusiasm ≠ user needs
Although the auction attracted nearly 40,000 onlookers, the number of bidders was very small. This shows that humanoid robots are still more of a "public topic" at present, and they have not really stimulated the desire to "buy and own". Curiosity can be watched, and decisions still need to be practical.
2. The threshold for operation is too high
Although the N2 has excellent mechanical structure, it requires some programming knowledge to operate. Without a graphical interface or natural language interaction, it remains a "difficult to use" machine for non-specialists.
> A simple question is: If the average consumer bought it at home, would it use it?
The answer may be disappointing. To make it run, turn, and stand firm, it must be controlled by complex programs, which also limits its consumer-level product attributes.
3. Practical application scenarios are still lacking
It can't stir-fry, can't mop the floor, can't chat, and is more of a "technological decoration" at home. Without a clear available scenario, it is naturally difficult to trigger a consumption boom.
Ⅲ The reality of the industry: similar to "people", intelligence is still far away
Today, humanoid robots are close to human form in appearance, but their behavioral intelligence is still in their "infancy".
The main current issues include:
Figure: Problem sorting out humanoid robots
This also illustrates the core pain point of the humanoid robot industry: the robot hardware has "run", but the software intelligence is still "slow".
Ⅳ The next step: how to make the robot really "into the family"?
In order to truly enter the civilian market, humanoid robots need to go through the following "last mile":
✅ 1. Natural language + multimodal interaction
Like talking to Siri, ChatGPT, use your voice to control bot behavior, and even interact with it through multimodal input such as gestures, expressions, etc.
✅ Second, modular functional application
Let the robot easily switch roles (greeter, elderly escort, sports coach, home safety assistant, etc.) like a "mobile app".
✅ 3. Integration of software and hardware + cloud support
Through local execution hardware + cloud computing power + large model support, the robot has both action ability and cognitive ability.
✅ 4. Situational self-learning
Through repeated practice and human guidance, the machine can form a "memory mechanism" to truly "teach, remember, and change quickly".
These will become the key breakthrough directions of the industry in the next 3-7 years. Once "teaching robots" becomes possible, humanoid robots will no longer be the darlings of the laboratory, but intelligent entities that truly serve the lives of the masses.
Ⅴ Conclusion: An auction, a reflection
This robot half-marathon is not only a science and technology show, but also a reality test. And the runner-up robot that was auctioned has become an "industry mirror" - reminding us that there is still a long way to go before "robots enter the homes of ordinary people".
But that's not bad news.
It means that there is still huge room for development in the market, that entrepreneurs and engineers still have sufficient innovation directions, and that the real era of human-machine symbiosis has not yet arrived, but it is not far away.
Technical cold thinking may be the most valuable sobriety in the craze.
