Figure 02 Released, Is the Future of Humanoid Robots Really $10,000?

Observer.com

2024-08-09 09:35Posted on the official account of Shanghai Observer.com

A few days ago, Figure AI, an embodied intelligence startup in United States that has been invested by Amazon founder Bezos, Microsoft, Nvidia and OpenAI, released its second-generation humanoid robot Figure 02. According to reports, Figure 02 has obvious improvements in design and functionality. After the release of Figure 02, founder Brett · Adcock also "called" Musk on social media.

This generation of robots is about 1.68 meters tall, weighs 60 kilograms, and currently moves at a speed of 1.2 meters per second. It is claimed that the new generation of dexterous hands has 16 degrees of freedom and a load capacity comparable to that of a human, and can grasp objects up to 25 kilograms; Powered by a custom model trained in collaboration with OpenAI, Figure02 can interact with people through built-in microphones and speakers, and 6 RGB cameras are embedded in the head and front and rear torsos, and the data from the cameras will be passed through a visual language model to help it with semantic understanding and rapid common sense visual reasoning. Compared to the previous generation, Figure 02 has 3x more on-board computing and AI inference capabilities; The Figure 02's battery pack capacity has also been boosted, with a 2.25 kWh battery pack that provides 50% more power than the Figure 01, and it is officially said to run for 5 hours, eventually hoping to enable it to work up to 20 hours a day.

Figure 02人形机器人

Musk, who was "provoked", is quite optimistic about the development of the field of humanoid robots, and said that Tesla will become a leader in the field of humanoid robots, and has also mentioned that the price of its robot products may be as low as $10,000, much lower than the price of cars, if each price is $20,000, Tesla can make a profit of $1 trillion a year.

In this regard, some industry experts believe that Musk is "bringing rhythm", he told the observer network that Musk wants other countries to go the route of low-cost humanoid robots, "continue to follow technology, rather than high-end manufacturing, we have to be wary of this kind of thinking." ”

Although the Figure 02 has been significantly improved in many aspects compared to the previous generation of robots, the industry is still in its early stages, and there are still some challenges in the future before robots are massively involved in production and into the home.

Judging from the domestic situation, the relevant start-ups are still very far from the real commercialization stage. According to the report compiled by Robot Headlines, according to incomplete statistics, the mainland is the country with the largest number of humanoid robot ontology manufacturing enterprises, with more than 60 companies, of which start-ups account for more than 50%. This means that the mainland has the most humanoid robot ontology start-ups; And there are zero real commercial breakthroughs.

Overseas, Musk, who has been interested in humanoid robots in recent years, is also constantly adjusting his expectations for the large-scale commercialization of his robot Optimus.

In April, Musk said that Optimus was already on factory duty and would sell the robots by the end of 2025. At Tesla's shareholder meeting in June, Musk said that Tesla would become a leader in the field of humanoid robots, but in July this year, Musk said on his social media that it expects small-scale production next year, and it is used internally by Tesla, and "hopes" to mass-produce robots for other companies in 2026.

Ding Ning, head of the Guangdong Embodied Intelligent Robot Innovation Center and executive vice president of the Shenzhen Institute of Artificial Intelligence and Robotics (AIRS), believes that with the current level of technology and safety, robots are still a long way from entering factories and homes on a large scale.

Regarding the future development of the industry, Ding Ning said, "We should focus on how to reproduce human ability, rather than human form." According to Southern Finance and Economics, Ding Ning believes that the strength of human beings lies in the brain, that is, the ability to think and solve problems, as well as the ability to create and use tools, and robots should imitate this, rather than simply imitating the appearance of people. "It is necessary to combine the application scenarios, rather than blindly pursuing the 'humanoid'", "For example, a hand does not necessarily need 5 fingers, some scenes may only need 3, and some may require 8. ”

A robot is not a "product", but a "methodology" for the integration of software and hardware. From academia to industrialization, AI and robots are highly interrelated, and the level of "intelligence" is the key to whether robots can complete their jobs well. "Therefore, we chose 'embodied intelligence' in the naming, hoping to build a more complete industrial chain, including hardware and software."

The level of "intelligence" depends on the quality and quantity of data. Experts believe that the mainland can become an important player in the industry due to its advantages in terms of the integrity of the industrial system and manufacturing costs.

Ding Ning introduced that the mainland started late in embodied intelligence, and it is relatively backward from chips to computing power, to talents, to the ecosystem. But the mainland's biggest advantage is that it has a complete industrial system and lower-cost manufacturing capacity, as well as a more efficient national system.

He explained that compared with developed countries that transfer low-end manufacturing, the whole industrial system allows the mainland to have a more complete data dimension, covering the physical world. Second, the mainland's well-established manufacturing system can make data acquisition cheaper. The robot is the interface and carrier of data, in addition to the "body" of the robot, it also needs to have production equipment and various molds. If manufacturing costs can be controlled, the cost of acquiring data will be reduced accordingly. By making good use of these advantages, we can obtain data more completely, cheaper, and more efficiently, and build a basic model of embodied intelligence.

This article is an exclusive manuscript of Observer.com and may not be reproduced without authorization.

View original image 188K

• 

• 

•