JAKARTA - The humanoid robot industry is expected to reach the "ChatGPT moment" by the end of 2028. The term describes the moment when a technology that was previously only known to a limited circle suddenly becomes popular, easy to use, and its benefits are immediately felt by the wider community.
In the context of humanoid robots, this stage means that robots are no longer just appearing in laboratories or exhibitions. The technology is beginning to be able to carry out various real tasks well and its usefulness is easy for the public to understand.
China Daily quoted Thursday, July 23, saying the projection was delivered by the founder and Director of Technology of Galbot Co Ltd, Wang He. The estimate is based on the rate of data collection and the tendency of convergence of models, i.e. the integration of various AI capabilities into one system.
According to Wang, a strong pre-training model and efficient follow-up training are key. A basic model of artificial intelligence in the form of a trained large and complete data, said Wang, has the potential to achieve a success rate of 70 to 80 percent when performing tasks that have never been specifically trained.
Artificial intelligence in the form of AI that works through physical devices, such as robots, to recognize the environment and take action.
This ability was seen at the World Artificial Intelligence Conference 2026 in Shanghai. Galbot G1 prepared breakfast, baked bread, and poured drinks in front of visitors.
The audience then moved the bread, shifted the kettle, rearranged the cups and plates, and closed the cups with their hands.
Galbot G1 does not stop. In a matter of milliseconds, the robot rediscovers the object's position, rearranges the motion path, and completes the task without interruption.
Wang said robots need an integrated system that works like a "brain and a cerebellum". The model must be able to understand the task while controlling the movement.
According to Wang, the technology is one of the paths to push humanoid robots towards the turning point of artificial general intelligence. This term refers to AI that is capable of understanding and working on various tasks, not limited to one function.
Production of humanoid robots in China has also surged.
China's Deputy Minister of Industry and Information Technology Ke Jixin said production reached around 20,000 units last year. That number exceeded 40,000 units in the first half of this year and is expected to reach 100,000 units throughout the year.
If the projection is achieved, this year's production will be five times that of last year.
However, the increase in production does not automatically mean that the industry is mature.
The Chairman of the Technical Committee for Standardization of Humanoid Robots and Artificial Intelligence and Chairman of the OpenAtom Foundation, Xie Shaofeng, said the industry is growing rapidly, but the perpetrators are still going it alone.
According to Xie, there are three main challenges. The technology lines between companies are very different, product standards are still lacking, and the demands for sustainable development are increasing.
Ke estimates that the next two to three years will be an important period for the sector. The industry is beginning to move from prototypes and technology demonstrations to the delivery of large quantities of products.
New problems arise when robots start to be produced in tens of thousands of units. Component mismatches, differences in testing methods, and unclear safety limits can grow into issues of cost, efficiency, and market confidence.
Therefore, scaling up is not just about increasing production. Components from different companies must be able to work together and the performance of robots needs to be comparable.
Standards are also needed so that users are confident in buying and have a reference for maintenance after the sale.
Ke asked for the drafting of standards for the classification of robot intelligence levels, data management, and safety to be accelerated.
Deputy Chairman of the Technical Committee Jiang Lei said hardware needs standards, while AI models need a "safety wall". The first group of standards is targeted to be issued by the end of this year.
The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)
