news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Mingmin from Aofei Temple
Quantum Bit | Public Account QbitAI

How popular is embodied intelligence?

From WAIC to MWC, various exhibitions, forums and press conferences,Humanoid Robot、Robot DogEmbodied intelligence has become an essential element. Wherever they are, people will gather, and they are well-deserved traffic codes for exhibitions.

Eye-catching, of courseMoney attracting。

This year, there have been continuous news of financing for embodied intelligence startups, and large orders have frequently been issued. Yushu Technology received 1 billion yuan in B2 round financing in February, and Galaxy General even received 700 million yuan in angel round financing.

The big model wave is still advancing rapidly, but the next protagonist leading the AI ​​wave has been booked by the market in advance.Embodied Intelligence。

However, just as there were many doubts when big models first became popular, the same is true for embodied intelligence today.

What is the path to achieve it? What is the prospect of commercialization? When will the outbreak come? What consensus and non-consensus have been formed now?

Just recently, Quantum Bit DialogueWang Guangxi, Lenovo Capital, and gained a glimpse into more understanding from the investment market.

Wang Guangxi’s view on embodied intelligence is simple and direct:

China has great opportunities.

“Embodied intelligence is like a smart car”

Who is Wang Guangxi?

Vice President of Lenovo Group, Partner of Lenovo Capital and Incubator. He built Lenovo's emerging business investment and spin-off system, and successfully incubated and spun off SHAREit, known as the "king of going global", and more than 10 excellent subsidiaries from Lenovo. He is responsible for venture capital in the fields of chip semiconductors, enterprise services, cloud computing, robotics, AR/VR, etc., and successfully led the investment in nearly 100 projects such as Smartsens, Yunji Technology, and Houmo Intelligence.

In the field of robotics, Lenovo Capital has invested inMore than 20, covering traditional industrial scenarios and emerging humanoid robot fields.

In the past year, Wang Guangxi has made frequent moves in the field of embodied intelligence. The companies he led investments in, such as Zhuji Power, Star Era, Kuawei Intelligence, and Benmo Technology, are now the darlings of major exhibitions.

Regarding the embodied intelligence track, Wang Guangxi and Lenovo Capital have shown a very clear attitude:

You need to invest. But you have your own criteria for which ones to invest in.。

The reason for investing is very straightforward: there are great opportunities in embodied intelligence in China.

Wang Guangxi believes that the field of robotics is very similar to the field of automobiles. In the era of traditional robots, just like the era of gasoline cars, Europe, the United States, and Japan have built a complete industrial system, ecology, and even industry standards. The specifications of various process products, underlying parameters, and core components are difficult for China to surpass.

To put it bluntly, 50 years of experience is at stake.

Correspondingly, embodied intelligence will become China's realization ofOvertaking on a curveopportunities.

Thanks to the development of artificial intelligence, algorithms can now be used to fill the gaps in hardware, such as using path planning algorithms to make up for the lack of sensor accuracy and using algorithms to optimize path planning.

There may also be changes in the architecture, and the traditional model needs to be abandoned, so that both domestic and foreign countries will be on the same starting line again, which is somewhat similar to the logic of smart cars.

In specific production and manufacturing, robots involve complex system integration work, involving chips, hardware, software control, applications, etc. Chinese manufacturers have always had their own unique advantages in experience and capabilities in integration.

In addition, in terms of market,China's interest in embodied intelligence started earlier。

In the past year, nearly 100 robot companies have emerged in China, and the primary market is extremely hot. Data shows that in 2023, the amount of financing in China's primary market robot industry reached 24 billion yuan, of which there were about 4 investment events with a scale of more than 1 billion yuan.

Relatively speaking, the latest wave of large-scale financing in the field of embodied intelligence overseas came later. In March this year, Figure AI, a humanoid robot company, announced the completion of a new round of financing of US$675 million, with investors including Bezos, Nvidia, OpenAI and Microsoft.

Recently, Skild AI, a humanoid robot startup headquartered in Pittsburgh, announced the completion of a $300 million Series A financing round, with investors including Bezos, Sequoia Capital, Japan's SoftBank, etc.

There are two reasons for the difference:

• First, the Chinese market itself prefers to invest in hardware;
• Second, robots are widely used in China.

Wang Guangxi said that from a regional perspective,China's scenes are the largest and richest。

According to the data from the Ministry of Industry and Information Technology, China's industrial robot installations accounted for more than 50% of the world's total in 2022, firmly ranking first in the world. The density of robots in the manufacturing industry reached 392 per 10,000 workers, and service and special robots were applied on a large scale in logistics, medical care, construction and other fields.

At the same time, China also provides a better development environment in terms of policies and regulations. The "14th Five-Year Plan for the Robot Industry" clearly mentions that the depth and breadth of robot applications should be further expanded. User units and robot companies are encouraged to jointly carry out technical tests and verifications, support robot manufacturers to implement key component verification, and enhance the test and verification capabilities of public technical service platforms.

This has made domestic corporate users more willing to cooperate and more willing to accept joint development with robot manufacturers.

Wang Guangxi believes:

For the robotics industry,China should become the largest user and producer at the same time., and it is even possible to become the most technologically advanced region in the future.

Based on these judgments and years of investment experience in the field of robotics, when the big model trend was in full swing, Lenovo Capital further focused its attention on the field of embodied intelligence. After scanning almost all startups in the field, it finally invested in 4 companies:

• Zhuji Power
• Cross-dimensional intelligence
• Star Era
• Benmo Technology

Why were these four selected?

Let’s first look at their actual progress. At Lenovo Capital’s annual Venture Week, these furniture-based smart startups all made their debut.

in,Zhuji PowerThe bipedal robot has achieved zero-sample, unprotected, fully open walking tests in the wild and can withstand strong blows.

During WAIC, it became an Internet celebrity because it was "bullied by humans."

Star EraThe humanoid robot can perform a series of human body movements based on remote operation, such as shaking hands, saying hello, etc.

At the same time, large models are also used to improve the actual perception of the physical world, such as vision, touch, and hearing.

Cross-dimensional intelligent robots emphasize grasping capabilities。

It can grab objects placed randomly on the table and can accurately pick up broken glass fragments. You know, glass fragments are irregular surfaces of various shapes and have their own reflective properties, which puts higher requirements on the robot's recognition and grasping.

It uses Sim2Real technology to train the robot arm operation based on 100% synthetic data, ultimately achieving 99.9% stable operation and grasping.

Wheeled and legged robot by Benmo TechnologyThrough innovative direct-drive joints and hub motor drive technology, the speed and agility of wheeled AGVs and the mobility and obstacle-crossing performance of legged robots are combined, enabling tasks to be completed efficiently and flexibly in more complex terrains.

It is understood that through modular structure and open interface, this wheeled and legged robot can be equipped with visual modules, communication modules, AI host, etc., and can be applied to smart parks, mines and other scenarios.

In summary, these four startups have some commonalities:

• The founders and team have strong academic backgrounds;
• Possess differentiated core technologies.

Lenovo Capital and Incubator values ​​these points.

Wang Guangxi said that with so much investment now, the startup market is inevitably mixed, and everyone thinks they can build a robot, or even make it walk a few steps. However, those who really want to do this industry, whether they are customers, R&D or investors, are more concerned about the most distinctive technical differentiation and underlying strength.

It is simple but meaningless to just make a humanoid robot.

For example, Zhuji Power has accumulated more than ten years of experience in combining sensory motion and algorithms. Wang Guangxi believes that Zhuji Power has a strong grasp of the motion control technology of robot quadrupeds and humanoid bipeds. Especially after the introduction of the reinforcement learning module this year, it is very important to solve the generalization ability of robot motion.

The research and development of humanoid robots by Xingdong Era can be traced back to the team's work at Tsinghua University. The algorithms are all developed by the team itself, and the research direction of founder Chen Jianyu has always been focused on the field of AI + robotics. Therefore, in terms of underlying technology, Xingdong Era has a strong "nativeness".

At the same time, talent itself is also an important variable. Young cross-border talents are more favored by Lenovo Capital.。

Wang Guangxi believes that embodied intelligence is a field that emphasizes systematicity and cross-border capabilities, and that requires R&D personnel to understand hardware, software, algorithms, and systems.

There has not been much overlap between talents in the software industry and the mechanical industry, especially the older generation of technicians, who may have a lot of path dependencies and their own areas of expertise.

This means that the embodied intelligence track requires relatively young, cross-border technical talents who are in the AI-native era.

In addition, the field of embodied intelligence itself is still a no-man's land, with many technical difficulties and high complexity, but at the same time, it has put the global academic and industrial communities on the same starting line, and everyone is crossing the river by feeling the stones.Young teams have more advantages。

In summary, with the right time, right place and right people, embodied intelligence has been assembled in China.

TimeJust as intelligence has reshaped the century-old automobile industry, the global robotics innovation force will also be put back on the same starting line because of intelligence. China’s opportunity to overtake others has come.

Geographical advantageChina has a vast number of robot application scenarios and markets. With the support and promotion of policies, the domestic acceptance of robots is higher, and the pace of robot + industry is also moving faster, which lays the foundation for the application of embodied intelligence.

People andWang Guangxi believes that the people who are working on big AI models and embodied intelligence are actually the same group of people. The New York Times reported that a study by the Macropolo think tank showed that, based on certain indicators, China has surpassed the United States to become the largest producer of AI talent, and has trained almost half of the world's top AI researchers. This means that China is training a large number of innovative talents in the AI ​​native era, who will become the core force driving the development of AI, embodied intelligence and other fields.

But opportunities and challenges often coexist.

As China ushers in a century of transformation in the field of robotics, challenges also follow.

The era of silicon-based intelligence has begun. Challenges? Opportunities?

First of all, from a technical perspective, the most direct problem isGeneralizationNot resolved yet.

Although strong AI represented by large models can provide a "brain" for embodied intelligence, there is still a long way to go before large models can be more intelligent and multimodal. For example, some modalities that existed less in the data world in the past (hearing, smell, etc.) have very little data volume.

At the same time, the "cerebellum" of embodied intelligence is not flexible enough. It is also a great challenge to allow robots to see the scene efficiently and smoothly, clearly understand the tasks to be completed, and execute them smoothly, safely and correctly. Even for humans, completing this process involves a set of complex biological mechanisms. There are still many technical difficulties for embodied intelligence to achieve such dexterity, generalization and intelligence.

Now everyone realizes that data is not as difficult as imagined.Through synthesis, imitation and other techniques, better quality training data can be generated.At the same time, piling up massive amounts of data is not necessarily the only path.

Secondly, in terms of manufacturing, start-ups often do not have their own factories and production lines. This is why everyone envies Tesla, because building their own factories can not only make them self-sufficient in production, but also accumulate engineering experience and innovation in production.

For example, Tesla has achieved one-piece die-casting from first principles, shortening the time to develop a new car to 18-24 months, while also sharply reducing costs. These engineering innovations, when put into practice and mass-produced, will be very beneficial to the application and deployment of Tesla's humanoid robots.

However, not every company has the ability or need to build its own factory, so it needs to find industry parties to provide it.

Finally, there are also ecological challenges in China.。

In the domestic To B market, users' willingness to pay is relatively low, and there is also serious internal competition among manufacturers.

This is even more true in the To C sector, where the rigid demand comes from the robot's generalization capabilities, scenario penetration and application services.

For ordinary consumers, buying a smart hardware is not only about buying the hardware itself and the software OS capabilities.It is also to purchase the ecosystem and various services built on the hardware., such as taking a taxi, ordering takeout, etc.

For To C, embodied intelligence also needs to build stable, secure, and efficient generalization capabilities, have service capabilities in a variety of scenarios, and build a rich ecosystem before it can gain consumers' willingness to pay.

Correspondingly, the field of embodied intelligence requires not only start-ups to continue to firmly carry out original innovation, but also the support of the industrial ecosystem and continuous investment in human, material and financial resources.

This is actually the commonality of hard-core original innovative technologies. For example, tracks such as AI, AR/VR, etc. all involve the construction and layout of complex ecosystems, and require long-term support and companionship. There is no shortage of such precedents in the industry. Lenovo Capital's history of investing in AI companies can be traced back to 2011. It has grown with them for a long time, and its ecological cooperation with Lenovo Group is also continuing to deepen.

Regarding the embodied intelligence track, Wang Guangxi also said that Lenovo Capital’s CVC model can provide manyLenovo Features's help.

First, Lenovo Group itself has a close relationship with embodied intelligence. Lenovo has large-scale 3C and electronic manufacturing smart factories in many regions in China, with rich scenarios. This can provide a "test field" for the implementation of embodied intelligence applications.

Secondly, Lenovo's intelligent manufacturing capabilities can provide industrial resources for embodied intelligent enterprises and solve problems such as early small-batch flexible production and experimental production.

Finally, Lenovo itself has rich overseas experience and resources, and its global layout over the past 20 years can provide local services and resources for embodied intelligent enterprises.

In fact, as a CVC, Lenovo Capital and Incubator has always emphasized empowering invested companies with the Lenovo ecosystem.

They proposed aThree-level ecological linkage system：

The first is the full-staff linkage layer. By establishing a member enterprise solution information database, Lenovo Capital and Incubator has established a daily docking mechanism, laying a good foundation for in-depth ecological co-construction.

The second is the resource empowerment layer, which is a resource empowerment platform that connects the entire chain of products, markets, customers, and platforms. This can help Lenovo Group to provide support to the invested enterprises, and the invested enterprises can also play a positive role in promoting the next development of Lenovo Group after long-term construction and development.

The third layer is the commercial implementation layer. Lenovo Capital and Incubator brings together the government, Lenovo Group, invested enterprises and industry customers to accelerate the implementation of technology empowerment in the industry.

Through these initiatives, Lenovo Capital and Incubator, with the help of Lenovo Group, has gradually formed a dynamic ecosystem.

However, as a CVC that is always committed to investing in original innovation, Lenovo Capital's investment will not be limited to the parent company's current business, but will plan for future innovation with a more forward-looking vision.Invest in the track of the next 5-10 years and invest in the greatest company of this era。

As a result, in the past decade or so, Lenovo Capital has invested in 18 listed companies, such asCATL,NIO,LotusThere are many industry leaders and unicorns in various vertical industries, covering multiple fields such as algorithms, computing power, autonomous driving, and intelligent manufacturing. The current layout of embodied intelligence is also based on the above thinking.

Regarding the latest trends, Lenovo Capital and Incubator proposed that the era of explosive growth of the silicon-based intelligent ecosystem is about to come.

The silicon-based intelligent era builds the future digital and intelligent world with the three major foundations of Code (AI) + Silicon (intelligent computing) + Body (intelligent body + hardware).

There are two key technologies:

• First, silicon-based intelligence driven by neural networks will surpass human intelligence through data training;
• Second, humanoid robots will acquire motor skills comparable to those of humans.

These advances will be driven by revolutionary advances in computing power and energy, making silicon-based intelligence pervasive and accessible.

Among them, investment and layout in AI has become a global consensus; and the investment in embodied intelligence and robots has just begun, and top scientists from all over the world are flocking to this field.

Wang Guangxi believes that the industry is still in its early stages.There is still a chance for great teams to step up and do the right thing.

The opportunities of embodied intelligence are self-evident.

Reference Links:
http://www.news.cn/info/20230818/e37e5e7233ac4048a02ddcc107210a88/c.html