Humanoid robots may be approaching a transformative moment similar to that experienced by generative artificial intelligence with the advent of ChatGPT. According to industry forecasts, this leap could occur within two to ten years, when these machines are capable of autonomously performing a large portion of the tasks given to them through voice or text commands.
The prediction was put forward by Wang Xingxing, founder and CEO of the Chinese manufacturer Unitree, one of the companies that has invested most heavily in the development of humanoid robots. According to him, the real turning point will come when a robot can interpret an instruction and execute it in an environment it doesn't previously know.
The idea represents a much greater challenge than simply creating machines capable of walking, running, or performing impressive movements. The goal is to develop robots capable of understanding their surroundings, making decisions, manipulating objects, and adapting to unexpected situations.
Wang estimates that an eventual “ChatGPT moment” could occur when a robot manages to execute approximately 80% of the received commands in a roughly 80% unknown environment. Currently, the technology is still far from this level of autonomy, mainly due to limitations in the generalization capacity of artificial intelligence systems.
The challenge lies in the brain, not just the body.
Humanoid robots are already capable of performing increasingly sophisticated movements. However, executing a previously trained task in a controlled environment is very different from being able to adapt to a completely new situation.
A small change in the position of an object, in the lighting, or in the characteristics of a surface can force the robot to readjust its movements. It is precisely this ability to generalize that continues to be one of the biggest obstacles to the development of truly autonomous robots.
Researchers are therefore working on artificial intelligence models capable of combining vision, language, and action. These systems will have to understand what a person wants, analyze the environment, and decide what movements are necessary to accomplish the task.
From the factory to the homes.
The first commercial applications of humanoid robots are expected to emerge primarily in industrial and logistics environments, where tasks are more predictable and can be controlled.
In the long term, however, the ambition is much greater. Manufacturers envision machines capable of assisting with household tasks, such as cleaning, transporting objects, preparing meals, or providing assistance with certain daily activities.
The founder of Unitree believes that, within a decade, a small robot could be in many homes. The technology could also be used in dangerous or repetitive jobs, including operations in industrial environments, hard-to-reach areas, or locations where there is a risk to human workers.
China leads the technological race.
China is currently at the forefront of humanoid robot development and production, with dozens of companies working in this area and strong government support for the industry.
Unitree has become one of the leading manufacturers in the sector, developing models intended for both research and commercial applications. The company has focused on relatively affordable equipment to accelerate adoption and enable universities, companies, and programmers to develop new applications.
International competition is, however, increasing. American companies such as Tesla and Figure are also developing humanoid robots for industrial applications, while other technology companies are seeking to develop artificial intelligence systems capable of controlling these machines.
There are still many obstacles.
Despite advancements, humanoid robots are still far from replacing human workers on a large scale. Limitations include energy autonomy, manipulation capabilities, safety, speed of execution, and, above all, the ability to understand new situations.
Another important challenge is the amount of data needed to train the systems. While language models can learn from the vast amounts of text and images available on the internet, robots need physical data: such as how to grasp an object, apply force, maintain balance, or react when something doesn't go as planned.
Creating this data requires real-world testing and thousands of hours of training, making development much more complex.
A new stage for artificial intelligence.
If the projected advancements materialize, humanoid robots could represent a new stage in the evolution of artificial intelligence: the transition from systems that operate primarily in the digital world to machines capable of acting physically in the real world.
The economic potential is enormous, but questions also arise regarding security, privacy, employment, and the misuse of technology.
For now, the "ChatGPT moment" for humanoid robots remains a prediction. But the speed at which technology is evolving suggests that the arrival of machines capable of understanding instructions and acting autonomously in human environments may be closer than it seemed a few years ago.
Content published on the Senso Magazine website is automatically translated using Google Translate services.
Publish it yourself!