Robots clean up welts, pick up cans in a directional manner, and even understand pop culture memes in the human world, and dance a "seaweed dance" in front of you...... The scenes in these sci-fi films are being accelerated to the real world by large AI models.
Before the advent of large AI models, robots had very limited thinking capabilities, and could only be limited to commands and actions, and could not understand natural language. However, with the breakthrough of AI large model products represented by ChatGPT, the capabilities of traditional robots have gradually been "unblocked". By implanting the "brains" of large models into robots, they can not only pick up your natural language, but even understand your humor.
Robots in action.
This brings the "Embodied Intelligence" that has been talked about in the academic circle for more than ten years into reality and becomes one of the hottest tracks in 2024. Some institutions predict that the global embodied intelligence market size will be $1.8 billion in 2023 and is expected to reach $13.8 billion in 2028.
In 2024, "accelerating the development of new quality productivity" will be written into the government work report and listed as one of the top ten tasks of the year. The government work report proposes to give full play to the leading role of innovation, promote industrial innovation with scientific and technological innovation, accelerate the promotion of new industrialization, improve total factor productivity, constantly shape the development of new momentum and new advantages, and promote social productivity to achieve a new leap.
In fact, every exploration of embodied intelligence enterprises is the excavation of new quality productivity, and they bring the capabilities of general artificial intelligence from the virtual world to the physical world, and then into thousands of industries.
However, in the process of building new quality productivity, many embodied intelligent enterprises are also facing problems such as organization and management, knowledge flow, etc., how will they keep up with the rapid development of technological changes, and constantly refresh their own productivity while bringing productivity improvement to traditional industries?
Let the robot "grow brains", and the embodied intelligent track is on fire
With the halfway point of 2024, which is known as the "first year of large-scale model application", the new large-scale model landing direction - the embodied intelligent track is out of the circle. According to incomplete statistics from China Robot Network, a total of 134 financing events will occur in China's robot industry in 2023, including 52 financing events of nearly 100 million yuan and over 100 million yuan, and the total financing amount in 2023 will be about 20 billion yuan.
Among them, the hottest is the Zhiyuan Robot founded by Huawei's genius boy "Zhihuijun" after his resignation. In the past year, the company has financed 6 rounds, and the list of investors includes Sequoia China, SAIC Investment, etc., and its valuation has soared to 7 billion yuan.
"Embodied intelligence" refers to the ability of an intelligent system or machine to interact with the environment in real time through perception and interaction.
Wang He, an assistant professor at Peking University and a scholar of embodied intelligence, once mentioned that embodied intelligence relies on the physical form and ability of individuals, and the lack of a large amount of ready-made data for training is the reason why the track has been in a bottleneck. But with the breakthrough of the AI large model, the large model allows the robot to begin to "grow brains". Industry practitioners have even begun to feed human motion data to embodied robots, allowing them to mimic human movements.
The cleaning robot is at work.
This means that in the past, the robot could only perform tasks such as sweeping and welding, but after adding the "brain", the robot can understand your humor, hand you something, clean the target garbage in a targeted manner, and even help you drive.
For example, the Aglitiy robot, which has been deployed to carry boxes in an Amazon warehouse, can not only disassemble how to cook on its own, but also understand human pop culture memes (such as the Sword of the Dark Lord Darth Vader in Star Wars) and learn to walk on roads that have never been trained to walk. Jacobi, a domestic robot manufacturer, has also released an embodied intelligent supermarket service robot, which can allow the robot to complete tasks such as replenishment and picking through verbal instructions.
How can AI scientists transform into sweeping robots?
Cutting-edge technology enterprises represented by embodied intelligence have become the focus of the industry, but after taking the lead, enterprises are always faced with the needs of intensive product changes and rapid business growth, and at the same time, they cannot avoid the challenges posed by rapid development to organizational management and business agility, as well as digital problems such as knowledge management and project management brought about by industry characteristics such as high knowledge density and high R&D intensity.
Recently, Zhejiang Youlu Robot Technology Co., Ltd. (hereinafter referred to as "Youlu Robot") completed an investment of more than 100 million yuan from Innovation Works, Yuanjing Capital, Baidu Venture Capital, etc., and won an order of more than 10 million yuan for the embodied intelligence universal brain Master2000, which fired the "first shot" of the commercialization of embodied intelligence. However, before this "gun" was issued, the company was once stumped by problems in organizational management and hardware manufacturing.
Youlu Robot was founded by Dr. Chen Junbo, the former technical director of the Damo Academy Robotics Laboratory and the general manager of the Damo Academy Xiaoman Donkey Company, and the most important product is a general robot brain, which can be placed on traditional professional equipment such as sweeping robots, logistics vehicles, industrial forklifts and forklifts through "plug and play" or simple transformation, and turn these equipment into unmanned equipment.
At first, Youlu wanted to sell software, that is, sell the part equivalent to the "brain" of the robot to traditional sweeper companies, but no company was willing to pay for an invisible and intangible "brain". After this model was trapped, the second idea of the Youlu team was to materialize the software and make a visible and tangible hardware, that is, a brain with an embodied shell, which could allow the robot to perceive and make decisions autonomously after insertion. "Customers generally felt that their eyes were shining at the time, but they were still not willing to pay." Gu Zulin, co-founder of Youlu Robot, said that the two encounters forced Youlu Robot to go directly to the sweeper.
So "Party B becomes Party A", Youlu Robot purchased a batch of sweepers from customers, integrated the "brain" into it, and launched a robot product called AI130, which can clean up the welt and accurately pick up the Coke cans on the road, and once launched, it created about three million revenues. This move proved the market value of the embodied intelligent universal brain, and also made the customer decide to buy the "head" model that Arilu originally peddled to him.
Cleaning robot AI130 in the Asian Games logistics warehouse.
Looking back on the process of moving from R&D to market, Gu Zulin said that the number one problem encountered was how to persuade customers to place orders, and the second major problem was process management. "The core team of Youlu Robot is AI scientists, and suddenly they want to engage in manufacturing and make sweepers, and everyone is not adaptable. First of all, it is necessary to go to a lot of production management systems and management ERP, including purchase, sale and inventory, etc., and suffer a lot of losses in the production management process. Gu Zulin said that the deployment of a system can be as little as two or three million, and as much as two or three million, and it is difficult for any entrepreneurial team to bear such a huge cost of trial and error.
In addition, for enterprises in the high-speed development track, practitioners often face the challenge of high knowledge density, and the rapid connection of data and experience is particularly important. Recalling this process, Gu Zulin was miserable. Taking the primary R&D stage of AI130 as an example, there are usually deer who will send multiple finished vehicles to work in the experimental scene, and will encounter problems from software and hardware such as lights not turning on, suddenly stagnation and not running, etc., engineers can adjust the problems on the spot, but when the vehicle actually runs, different vehicles will have various abnormalities in different states, and the operators cannot pull together at all.
In order to solve the above problems, Youlu introduced the DingTalk enterprise low-code platform Yida, in which the process of each maintenance and the problems that occur are collected and sorted out one by one, and the follow-up product team will convert every subtle problem into a work order. According to a unified problem sheet, the above-mentioned tests in multiple experimental sites and different conditions of multiple vehicles can be solved in comparison.
"Different stages of enterprise development have different requirements for software, before gritting teeth in the past, or let it develop in chaos, after the introduction of low-code, our internal management has reached a level, including R&D fleet management, data platform can be opened, and the internal flow has become very silky." Gu Zulin said.
Low-code, collaborative documents, and intelligent Q&A inject new productivity into the enterprise
Coincidentally, another Shenzhen-based mobile robot and solution provider, Youai Zhihe, also encountered similar problems. As a national specialized and new "little giant" in the Greater Bay Area, Youai Zhihe applies the core technology of high-precision SLAM navigation mobile robots and software systems, and provides autonomous mobile robots and integrated solutions for global enterprises based on the two major businesses of industrial logistics and intelligent inspection and operation and maintenance.
Around 2022, the management team of Youai Zhihe found that with the continuous increase in business volume, the more urgent need for information management is gradually approaching, because the management process is not standard, and the way of "hand-in-hand" has not been able to bring about the overall development, and the management difficulty has gradually emerged. In order to solve the above problems, in 2019, Youai Zhihe began to apply DingTalk, and in 2022, it will be used in depth. According to reports, with the help of low-code MES production execution system to achieve production and manufacturing management, Youai Zhihe has saved millions of procurement budgets.
The MES production system built by Youai Zhihe low-code.
In addition, Chang Weijia, head of IT technology at Youai Zhihe, observed that with the limited number of technical support team members, with more and more enterprise projects, the demand for intelligent answers to some general and repetitive questions is becoming more and more vigorous. In this regard, Youai Zhihe chose to apply the AI capabilities on DingTalk to set up an intelligent Q&A robot to achieve rapid connection and learning of industry knowledge and professional knowledge.
At present, Youai Zhihe has launched the AI robot intelligent question and answer function, which can take the lead in using the DingTalk robot to answer technical problems when the on-site deployment employees encounter technical problems, saving the technical and labor costs of the headquarters. In the future, Youai Zhihe expects to solve 60%-70% of the questions on it.
This coincides with the "document culture" of Youlu Robot. According to reports, Youlu Robot has advocated the weekly meeting rule of "no document, no meeting" since the first day of its establishment. "The rule of our weekly meeting is to list the documents first and then hold the meeting, and the problems encountered in the process of the meeting can be directly in the document, and then converted into a to do list into the corresponding employee's schedule, which is a very efficient collaboration and management tool." Gu Zulin introduced. At present, Youlu Robotics Co., Ltd. can realize collaboration based on knowledge base documents from product development to production, so as to improve the efficiency of R&D and testing management such as product testing and restructuring.
"Every valuable word that our company has said since its establishment is in the document, the worthless words are in the chat record, the words spoken offline are all nonsense, and the process is all in Yida." Gu Zulin teased.
Looking to the future, although humanoid robots with similar behaviors to humans are one of the directions that have attracted much attention in embodied intelligence, the application of embodied intelligence in industrial, logistics and other scenarios, and the combination of various equipment is its broader value. Youlu Robotics, Youai Zhihe and other embodied intelligent cutting-edge enterprises that have grown up in the AI wave, specialized and new "little giants" will undoubtedly enter all walks of life and become new productive forces in the AI era, and AI productivity tools represented by DingTalk are also an indispensable part.
Recently, the "2024 China's Top 500 New Quality Productivity - National Economic Growth Potential Report" jointly released by the financial information big data platform Feidi Technology, Feier Data and Communication Research Institute, and Yicai Business School shows that more than 60% of unicorns and cutting-edge breakthrough enterprises are using DingTalk, including many new scientific and technological forces including AI and embodied intelligence tracks. More and more enterprises are using DingTalk to reconstruct production relations and cultivate new quality productivity.
Producer: Nandu Political and Business Data News Department, Shanghai News Department
Written by: Nandu reporter Lin Wenqi and Ma Ningning