"At first, we thought this was an opportunity for the Internet to not meet in a decade, but the more we thought about it, the more we felt that this was an opportunity similar to the industrial revolution that invented electricity that was not encountered in hundreds of years." On May 18, at Tencent's 2023 shareholders' meeting, Ma Huateng, chairman and CEO of Tencent, expressed such feelings.
"As we all know, our human race has experienced three industrial revolutions, each of which has brought new developments in science and technology, and has also had a profound impact on social production and lifestyle." Wu Qilin, chairman and chief technology officer (CTO) of Elephant Robot, said that it is now in the fourth industrial revolution, represented by cloud computing, Internet of Things, big data, robots, etc., is promoting the change of social production and lifestyle, especially the recent rise and rapid development of artificial intelligence (AI) large models, which has further promoted the development of robots, especially humanoid robots.
At present, AI technology is gradually and deeply embedded in all aspects of China's industrial production and high-end manufacturing. Industrial practitioners have called it a "moment of excitement".
Shao Tianlan, founder of Mecamand Robot, is also very excited about the application of AI large models in industry. "Now the large model gives a very good idea, it has hundreds of billions of parameters, through a variety of training data across modalities, when the basic capabilities are done, the large model can produce a more advanced ability." He pointed out, "This ability can replace the work that many engineers can do, and by combining this ability, we think there is an opportunity to push the technology of robotics to another order of magnitude, which is also a huge opportunity that we are seeing now." With the big model, the market will be 10 times or even 100 times the existing one, and that's something we're very excited about right now. ”
Bionic robot business consulting doubled
"Big models drive a whole new wave of the entire AI industry. Nowadays, all walks of life are exploring the combination with large models. Liu Wei, CTO of iFLYTEK Cloud Platform Business Group, said at a recent "iFLYTEK AI TechDay" event that as the most core and comprehensive application carrier of artificial intelligence, this combination between robots and large models will inevitably play an important role in promoting our productivity.
"From the data of our platform, we can see that in the past six months, the call volume of our entire service robot has increased by 88%, and the number of bionic robot business consulting has increased by 125%. What is more gratifying is that in the past, robot applications may be more concentrated in some service areas, and now we can see that robot applications appear in industry, manufacturing, agriculture and even logistics industries, and there is a gratifying phenomenon of penetration into the whole industry. Liu Wei said.
The reporter saw at the scene that relying on AI technology and iFLYTEK's "robot superbrain platform", a crawler robot can quickly build motion and navigation capabilities: after a series of exploration and map construction, the robot can easily achieve a navigation task from point A to point B, and also has real-time position display and navigation and obstacle avoidance capabilities.
The above cases are just the beginning. In fact, AI has long been applied in the field of industrial production. Recently, the reporter saw in the Shenzhen branch of a German company that the company's laser welding equipment has begun to use technology for some high-end equipment manufacturing. For example, in response to the problem of laser welding of copper wire in the motor, a laboratory leader of the company said, "We mainly use machine vision to judge, some copper wires with excessive spacing will not be welded first after artificial intelligence judgment, but they will also be marked one by one, and later by manual intervention; For some relatively large but acceptable gaps, after machine vision recognizes the special state, it will automatically adjust the welding parameters, including welding trajectory, power, etc., in short, try to make the equipment intelligently respond to various working conditions to complete welding. ”
"This equipment uses machine vision and artificial intelligence, in the early stage, through a large number of production pictures, and then manually to mark the correct part, after a certain amount of accumulation, the AI model can find out the law on its own to know which is what you really want, it will filter out all interference factors, and it can also assist in post-weld quality inspection." The head of the aforementioned laboratory said.
iFLYTEK also introduced a case of AI-enabled industrial production to reporters. In order to assist the digital transformation of the steel industry and build a modern production line, Shougang Qian'an Iron and Steel Co., Ltd. in Qian'an City, Hebei Province, used an iFLYTEK acoustic imager to visualize the leakage speed sweep of the vacuum of 12-way sealed branch pipes of the sintering plant, directly shortening the inspection time that used to take more than 10 hours to 90 minutes, and detected 22 gas leakage points in the plant, so that the location of gas leakage in the steel plant was "clear at a glance". If the annual output of the plant is about 1 million tons, this maintenance can save the plant 3 million ~ 5 million kWh (a total of about 900,000 yuan).
Improve the level of industrial automation
At the 6th Digital China Construction Summit in April, Daniel Zhang, chairman and CEO of Alibaba Cloud Intelligent Group and CEO of Alibaba Cloud Intelligent Group, revealed that Alibaba Cloud engineers are experimenting with connecting the Qianwen model to industrial robots, entering a human language in the DingTalk dialog box to remotely command the robot to work.
This also means that the birth of AI large models brings new possibilities to industrial robots. According to Alibaba Cloud engineers, large models can play an important role in robot development and application. In the robot development stage, engineers can automatically generate code instructions through the Qianwen model to complete the development and debugging of robot functions, and even create some new functions for the robot, such as the robot can arbitrarily combine and arrange basic capabilities such as grasping and moving, so as to complete more complex tasks; In the production run, the Qianwen model provides the ability to reason and make decisions for the robot to perform tasks, and front-line workers only need to send a piece of text, Qianwen can understand its intention for task reasoning, and automatically translate it into code that the machine can understand, command the machine to perform the task, and effectively improve production efficiency.
Wu Qilin said that there has always been a pain point in industrial production, that is, in robot programming and process debugging, the programming time is long and the deployment cost is high, but the AI large model is expected to solve this problem.
"For example, in the coating process, whether we can train through some models in the future, enter user parameters, it will automatically generate a set of robot trajectories and control process equipment, and then carry out some fine-tuning or mechanized debugging on this basis, so that the entire programming and application become simpler and faster." Next, the entire process improvement and process programming can be further deeply combined with the AI model. ”
Wang Xiaoling, managing partner of Times Bó Lè, pointed out in an interview with the Securities Times reporter that an important application of AI large models in the industrial field is to improve the level of industrial automation with the help of massive industrial data. "For a simple example, it can be of great help in the field of industrial manufacturing, first of all, in the field of machine vision, because machine vision has been limited by the entire computing power and the limitation of algorithms in the past, and there is no way to truly judge the behavior of an object autonomously, but more according to the established operation instructions." If the AI large model is connected to machine vision, it can efficiently complete some disorderly grasping and put it in the right position, which is a more typical application case. ”
Wang Xiaoling said that with the improvement of overall data computing power, AI large models can also implement learning-based regulation and control system (PNC) plans in unmanned factories. For example, according to the inventory history of customer orders, production scheduling and inventory, you can very accurately determine when to place an order and when to produce, instead of making production plans based on human experience in the past, so that the entire production efficiency has been greatly improved.
Change is quietly taking place
In recent years, with the continuous development and application of AI technology, industrial production and high-end manufacturing have gradually begun to realize intelligent transformation. More and more companies hope to reduce costs and increase efficiency and improve product quality by introducing this technology.
Policy emphasis on artificial intelligence and AI big models is also increasing. On April 28, the Politburo meeting of the Central Committee mentioned that it is necessary to attach importance to the development of general artificial intelligence and create an innovation ecology; On May 5, the 20th Central Finance and Economic Commission held its first meeting, which mentioned that strategic emerging industries such as the new generation of information technology are an important force to guide future economic and social development, and it is necessary to grasp the wave of new scientific and technological revolution such as artificial intelligence.
But Rome was not built in a day. Although AI models are expected to set off a new round of industrial revolution, this is a gradual process, not overnight.
"The speed of the real application of AI large models to industrial production and high-end intelligent manufacturing will be very fast, and the promotion may appear as rapid as the popularization of early smart machines." Wang Xiaoling said, but it has two bottlenecks to solve, the first is computing power, this large-scale computing power can be solved through the cloud at a lower cost. A more difficult point is data collection, because the overall degree of data in the industrial field is not high, and more human input is required. Visualizing and digitizing production operations is expected to take five to ten years.
Wang Xiaoling believes that for the traditional manufacturing industry, capturing these data requires a lot of sensors and microservers, combing and then uploading to the cloud, which is a huge upgrade. But this requires a large amount of capital investment by manufacturing companies to achieve, long-term profit accumulation, and private equity financing or listing support before it is gradually applied.
In the face of increasingly complex industrial production and high-end manufacturing, AI large models still need to overcome one specific difficulty after another. Han Zhuo, chief development officer of Mecamand Robotics, said: "I have visited about 200 factories in the past two years, and our characteristic is that we are always in the factory. In the field of AI+ manufacturing, everyone says that it is a trillion-level market, but from a developer's point of view, we feel that it may be more like a 10,000-level market, because there are a large number of non-standard applications. ”
"Many overseas industrial robot companies will find that they are not satisfied with the soil and water when entering the Chinese market, because management and production are standardized in the West, but in the Chinese market, there are a large number of non-standard needs for both management and factories." Han Zhuo said, "What we can do now is to face a variety of non-standard and trivial applications, which can be solved in a standardized way." ”
Responsible editor: Wan Jianyi
Proofreader: Zhao Yan
Blockbuster trailer: The venture capital circle is boiling! The 11th China Venture Capital Forum Kicks Off! Just on June 9
Super guest lineup, super industry alliance, super combustion investment topic...
Copyright Notice
All original content of each platform of Securities Times shall not be reproduced by any unit or individual without written authorization. Our company reserves the right to pursue the legal responsibility of relevant actors.
For reprinting and cooperation, please contact the assistant of Securities Times, WeChat ID: SecuritiesTimes
END
Click a keyword to view it
Periscope series in-depth report丨Stock Council column丨Investment Little Red Book丨e company investigation丨Times living room丨Top ten star private placement interview丨Sudden! 3 more companies delisted! 50,000 shareholders can't sleep丨11 board "monster stocks" burst out! The chairman of the board of directors and the chairman of the board of supervisors both resigned, and their annual salaries exceeded one million... 丨Smash tens of billions to expand production! Two hundred billion photovoltaic leaders announced on the same day丨Beijing found two cases of monkeypox, disease control emergency reminder 丨"One point is worth one yuan, college entrance examination points are worth special money", free red wine, set up a caring team... Wine enterprises fancy to help the college entrance examination! 丨The exam is open! How many points can you get in the 2023 college entrance examination A share paper?