The humanoid robots that pick up coffee, fetch apples, and spread pancakes are highly anticipated at the 2024 World Robot Conference.
During the same period, Fu Sheng, Chairman and CEO of Cheetah Mobile and Chairman of Orion Starry Sky, gave an exclusive interview to the Beijing News Shell Finance at the conference and expressed his pessimism about bipedal robots. He believes that such robots can be developed, but it is impossible to truly achieve commercial implementation within three to five years. Stability and cost are the core issues for its implementation
The form of a robot does not necessarily have to resemble a human, "said Fu Sheng. In his vision, the mainstream form of future robots is: one wheel, one arm, plus interactive voice and screen.
The robotics industry lacks a specific scenario that is truly well done
Beike Finance: In your opinion, how can AI (Artificial Intelligence) big models empower robots, and which skills are most likely to empower and enhance them?
Fu Sheng: A big model is a module that is based on language learning to understand the world and has its own logical and reasoning abilities. Its great empowerment for robots is that it can become a brain. Previously, robots essentially had no brain and were more automated machines. Firstly, large models will enhance the autonomous decision-making ability of robots. After robots have a brain, and the knowledge level of the brain is still good, various abilities are greatly improved, such as explanation, interaction, and task processing. In the long run, robots can truly participate in some labor, such as adding a robotic arm to the robot and letting it help you carry a cup of water. These should soon be commercialized.
Shell Finance: At the 2024 World Robot Conference, some robots can help people fetch coffee, brew coffee, spread pancakes, play chess, write calligraphy, and more. Where are the truly valuable scenarios for these robots in the future?
Fu Sheng: Many robot movements that can be seen today are not intelligent, but pre written code. If we can truly transition from prefabrication to intelligence and achieve good costs, it would be a huge scenario. I'm not worried about what the scenarios for robots are. This industry doesn't lack so-called scenarios at all, but rather lacks a scenario that is truly well done. These robots are really like humans running pancake stalls, able to face any unexpected situation, which is quite difficult today.
Beike Finance: There are almost no large robots in the 2024 World Robot Conference, more medium to large and some small robots, as well as robots in industrial application scenarios. Do you think this is an overall progress of robots?
Fu Sheng: Perhaps because there are relatively few foreign exhibitions, large robots that can lift large objects are mainly produced by companies like Fanuc in Japan. In fact, before the emergence of the big language model AI, the development direction of robots was high carrying capacity and high precision, which humans could not achieve. But the new direction is to make it intelligent, truly capable of achieving human level or even better than human performance in some scenarios.
The task that was previously considered difficult in the industry has now been achieved, but it should not be overly exaggerated. The production of bipedal robots in the workshop will not be realized for a long time, and the complexity is very high. What is the problem in the robotics industry? Everyone thinks that I want to make a super Rocket product. I will be able to do everything just like people. Then every family will have it. It won't be like this. Many products are gradually made good in a certain scenario and then gradually grow out. There has never been a super Rocket in the industry, which is difficult to achieve in terms of cost and complexity.
Beike Finance: So you're not particularly optimistic about humanoid robots, are you?
Fu Sheng: I am not optimistic about bipedal robots. I am optimistic about adding robotic arms to robots to do some work. For a long time, the stability and cost of bipedal robots have been the core issues for their implementation. I am not pessimistic about the development of bipedal robots. This direction can be developed, but it is impossible to truly achieve commercial implementation within three to five years.
Robots don't necessarily have to be humanoid in form, and their cost needs to be low enough
At the opening ceremony of the 2024 World Robot Conference, Academician Qiao Hong, on behalf of the organizers, released the "Ten Trends and Prospects for Humanoid Robots", which mentioned the ethics and safety of humanoid robots. How to ensure that humanoid robots do not exceed ethical and safety standards when serving humans?
Fu Sheng: When equipped with autonomous decision-making power, all robots have this problem. The core is still to keep up with management and legislation. Robots are tools, even with a large language model and a certain understanding of the world, they will not generate consciousness. But after the tool becomes powerful, who will use it and how to use it will be a problem. I support strengthening regulatory, legislative, and management measures to make it like the current big language model that requires dual product filing through the Cyberspace Administration's algorithm.
Beike Finance: In the trend of aging population, some robots focus on accompanying the elderly. Previously, an academician expressed the hope that language models could be empowered on robotic vacuum cleaners or have accompanying functions. How do you view the demand for robots in the context of aging population?
Fu Sheng: This is a huge opportunity. Companionship is indeed a big scene, and the reason why the previous companionship was not done well was because the technology did not reach this level, and products like Chat GPT did not appear. At that time, the novelty of companionship disappeared after a few words. The elderly do need more care, but in fact, they are the loneliest.
Beike Finance: Please outline the overall development trend of robots in the future.
Fu Sheng: At first, I imagined that robots would move from factories to the service industry, and eventually enter thousands of households. After the big language model was released, the robot had already reached the eve of an explosion. The form of a robot does not necessarily have to be like a human. It can truly help you complete a certain task, and the cost should be low enough. It is not achieved by acting like a human, but rather by better fulfilling its functions in order to meet our various needs. I think a mainstream form of future robots is: one wheel and one arm, plus a great interactive voice and screen.