Ai landing this thing is actually very simple, can not grow downwards and industrial resonance of AI, not good AI
Wen | Shiran
Winter moon, early morning. The water outside dripped into ice, but liu xin, a steel worker, was sweating like rain.
He pushed open the iron door of the workshop, and a cold wind like a dagger quickly turned into frost on his hot and wet eyelashes.
High temperature is the background color of day-to-day work on the blast furnace. White-collar workers in Beijing, Shanghai and Guangzhou are accustomed to a constant temperature in the office of 26 degrees, and in a steel workshop, Liu Xin's working environment may be as high as 62 degrees. The production of steel is non-stop day and night, as a quality inspector, Liu Xin's life has become "three shifts", and after the day shift, there are two consecutive night shifts.
Working in the workshop for decades, Liu Xin was on night shifts two-thirds of the days. In the airtight protective clothing, in the monotonous and silent quality inspection action, he gradually grew old.
In manufacturing, quality inspection workers are a special group. Their presence is important, the last hurdle for product quality, but they are less important. The gold content of the work engaged in for a long time is not high, the skilled work type, the upward space is almost non-existent, and the consumption of youth consumes the body.
A manufacturing company owner said to us, "What I value most is the equipment maintenance staff and the new product developers (and craftsmen), other people don't care, they can be replaced with a new one right away." ”
A senior executive at another equipment manufacturing plant said, "I have never felt that the quality inspectors are playing any role, except for a little intimidation of the workshop." "In an age as precious as gold, young people should do more valuable things, and this thing can be replaced by a machine." ”
When technology develops to a certain extent, this thing is not impossible, and it is becoming a reality. And it's not just about quality control.
value
The pain of the needle cone in the eyes appears from time to time, which is the trouble in the heart of Wang Lina, a quality inspection worker in the factories under Hengyi Group.
As one of the world's maximum fiber manufacturers, Hengyi Group's quality inspection workers, Wang Lina's job is to take a strong flashlight to check the products, every day, she needs to work in strong light for eight hours, manually inspect 2500 pieces of chemical fiber products. When she encounters a product in question, she also needs to manually hold it down, each weighing up to 15 kilograms.
In the past, workers held a glare flashlight to check the product with the naked eye, which was easy to cause damage to vision
Wang Lina has been engaged in quality inspection work for a shorter time than Liu Xin, but she has also been doing this line for 15 years, and the hidden worries of vision and physical decline are snowballing.
The efficiency and accuracy of quality inspection directly affect the production efficiency of enterprises. So as workers age, one day they will no longer be able to perform the jobs they once did. This once seemed to be the fate of Liu Xin and Wang Lina.
Fortunately, their fate suddenly changed.
Artificial intelligence has pried open a turnaround. This summer, Hengyi Group and Baidu Intelligent Cloud cooperation to create intelligent quality inspection equipment in Zhejiang Linjiang Industrial Park landed, for Wang Lina, since then, her job has become sitting next to the assembly line, looking at the products identified as "abnormal" in front of her, and then carrying out a round of manual re-inspection. If it is found to be abnormal, it is sent to the abnormal product area; if it is mistakenly detected by the machine, the product is sent directly to the packaging workshop on the assembly line by pressing the button.
Nowadays, with intelligent quality inspection equipment, the original general inspection has become a re-examination after machine screening
For Liu Xin, it only takes a set of cameras to match the production system to free him from the fire-baking environment - they are still in this precision-run production system, but they no longer need to couple their health and will to heavy labor.
Expanding the living space of people in the intelligent era, and even the value of people, this is the most simple role of industrial intelligence.
"When we really go to the enterprise, we find that the transformation that technology brings to the manufacturing industry is not the use of AI to drive digital transformation (it is over), this process has a long-term impact." Baidu Vice President Li Shuo said to us. He said that "AI + industry" not only solves not only technical problems, but more importantly, the core capabilities of manufacturing industry, and the leap in talent resources.
For a long time, China's industrial development has been insufficient in the underlying technology, and production optimization is often based on the extreme use of human efficiency, however, this way of organizing production is difficult to sustain today, and young people are generally reluctant to accept it, nor is it a sustainable development model.
A hidden contradiction is that when workers have the opportunity to get out of traditional work, what kind of changes will occur to their professional requirements and direction? Will they pursue new careers, or will they stay and apply their experience to the annotation and model training of AI data, and then invest in the future of AI?
At this time, if you look at this problem from a more macro perspective: when the industry is intelligent, it is already the consensus of all walks of life. What kind of path is suitable for transforming the seemingly heaviest industrial track? When the relationship between man and technology, technology and industry, industry and people changes, how do we find our coordinates again?
path
Throughout 2021, Li Shuo went to more than 100 factories; he also had a new understanding of industry, "Industrial production pays attention to seeking truth from facts, and results are the first." ”
This is very different from the era of the consumer Internet. Consumer Internet heavy model innovation, do not need heavy asset investment and independent technology research and development, wrapped in traffic and capital, you can spark.
However, the background of industry is slow and grinding, which makes the industrial track not sought after by capital, and it is even more impossible to attract talents and rapid development.
However, industrial transformation is imperative. It's a tough but necessary thing to do.
In 2012, the more than 100-year-old Kodak Company went bankrupt. In the "film era", Kodak's market share once reached two-thirds of the world's, but when digital technology came, Kodak's transformation was slow and suddenly fell. The "Kodak Moment" has since been used as a metaphor for "subversive death."
When the Fourth Industrial Revolution came, the clouds of the "Kodak Moment" hung over traditional industrial enterprises.
Industry is the foundation of the national economy and the stabilizer of economic development.
There have been three waves of transformation in industry: the first wave of digital manufacturing, which lasted from the middle of the last century to the mid-1990s, was characterized by computing, communication and control applications; the second wave of networked manufacturing, which has developed from the mid-1990s to the present, is characterized by the large-scale popularization of the Internet.
At present, big data, cloud computing, machine vision and other technologies are developing rapidly, artificial intelligence is gradually integrated into the manufacturing field, advanced manufacturing, and began to enter the intelligent manufacturing with a new generation of artificial intelligence technology as the core.
This also means that ai companies are currently entering the industrial track, such as flowing into the sea. Even if the sea does not only mean opportunities, there are unknowns and dangers.
In Guiyang, 2,100 kilometers from Beijing, Li Shuo ran three times.
For the implementation of industrial digital transformation, the larger the scale, the more obvious the network effect, and the more likely it is to form a group effect. Guiyang, the capital of the Yunnan-Guizhou Plateau, ranks 19th in the total GDP of the 27 provincial capitals (excluding municipalities directly under the central government and Hong Kong, Macao and Taiwan) in the country (2020 data), with a weak industrial base and a high degree of dispersion of enterprises.
Li Shuo told reporters that when he first went to a local company to investigate, the team had already implemented some digital projects for the company, however, the person in charge of production suddenly found that there was no network when showing him production data.
The workers seemed a little overwhelmed, but Li Shuo thought that he was the one who should be sorry, "This shows that we have not done enough." If the pain points of workers in production are really solved, they will use them every day, how can it happen that they forget to install the network? "The process of digitization should not stop at the superficial stage of "production data - large-screen display". In the following month, Baidu's intelligent cloud team successively landed the specific operations, MES, quality inspection and other links of the workshop workers one by one, from the production screen, the display of the workshop control center, and then to the piece income seen by the worker master through the mobile phone every day.
When Li Shuo walked into the factory "stealthily" again, he asked the master worker on the production line, "Will you use this scanner and the landing system?" "Yes, so I can see the number of finishes per day." Later, the company had a full-time network and information system administrator, and the paper documents that the production leader had to sort out every day became the numbers that jumped in real time on the screen.
Digitalization has changed from an abstract process to a step-by-step, solid reality.
A strong sense of mission for industrial intelligence runs through the entire process of Baidu intelligent cloud entering the industrial track.
The change begins in the second half of 2020. In October last year, Baidu Intelligent Cloud and the Guiyang Economic Development Zone Management Committee jointly established Baixun Zhihui Company, which is positioned as an industrial big data operation service provider and is currently the only industrial Internet team in Guizhou with the ability to integrate software and hardware research and development.
In Guiyang Economic Development Zone, the methodology of Baixun Zhihui can be summarized as two steps: horizontally facing the industrial park to expand the platform; vertically for specific industries to do deep benchmarking cases.
In 14 months, Baixun Zhihui initially built an industrial Internet service platform, some figures can explain its changes in the past year: bringing together 9 types of applications, 12 industry solutions, forming 15 benchmarking demonstration enterprises, preliminary investigation of 735 industrial enterprises in the economic development zone, through the industrial Internet will be each enterprise's equipment information, personnel information and degree of informatization into the background management, and cooperate with local enterprises to operate equipment maintenance service center, quality inspection service center, Marketing promotion service center, industry knowledge empowerment center and safety production service center.
Li Yingting, director of the industrial Internet department of Baixun Zhihui, believes that Baidu is a digitally native enterprise with natural advantages in AI cloud computing, big data and deep learning. Specific to the product, Baidu has the advantage of traffic, the second is the experience of creating mature big data products in the past; and the third is the advantage of the entire Baidu ecology.
Li Shuo said that Baidu Intelligent Cloud has a unique advantage in artificial intelligence, especially computer vision and knowledge graph.
But in industry, AI seems to have a hard time telling sexy stories.
Li Yingting gave an example to reporters. Two years ago, a local company called Jinjia Packaging also had a safety accident because the hair of female workers was caught in the production line. For specific scenario needs, they have made three types of applications for this packaging company.
One is the application of safe production, mainly in monitoring, combined with Baidu's AI capabilities, some algorithms such as tooling, work caps, detection of fireworks, dangerous chemicals between the forbidden, electronic fences and so on;
The second is energy consumption, for which more than 110 points have been made, mainly according to the power consumption of each equipment in the production workshop and the entire equipment switch and abnormal situation, online detection. Previously, enterprises were quantitative in electricity. For electricity consumption data, the company does not have a reasonable plan, but in fact, enterprises can not consume so much electricity, saving 30,000 to 50,000 yuan per month in electricity bills, and saving 400,000 to 500,000 yuan throughout the year. For a business, this is not a small expense.
Third, for the temperature and humidity of this piece, for the production of packaging workshop, temperature and humidity is particularly important. Here, Baixun Zhihui introduced information means to control, such as automatically adjusting the temperature of air conditioning and automatically adjusting air humidity.
Behind this, the integration of AI and industry is heavy and not easy.
Earlier, enterprises wanted to digitally transform, that is, "go to the cloud", deploy the business, process and application of the enterprise to the cloud, and now, "going to the cloud" is only the first step, the focus is on how to make the business, process and application run faster and better on the cloud after going to the cloud, so that the business intelligent upgrade. That is, "cloud and wisdom are one".
Li Shuo believes that the most valuable application of Yunzhi integration in the industrial field is that "we have run to improve the efficiency and quality of people's production at the beginning, which is the most core proposition, industrial production has its own beat, any technology can not slow down the beat, in the case of not slowing down the beat, if you can improve the beat, it corresponds to the efficiency of the entire production line can have a greater improvement." ”
In the past, it was to fight high, but now it is to accumulate steps and then talk about thousands of miles.
resonance
Seeking AI landing based on cloud computing is not a master key.
Today, China's cloud computing market is forming a "four clouds" pattern, namely Alibaba Cloud, HUAWEI CLOUD, Tencent Cloud, and Baidu Intelligent Cloud. According to the third quarter report of China's cloud computing market released by third-party data agency Canalys in 2021, Baidu Intelligent Cloud has the fastest year-on-year growth rate, becoming the only cloud vendor to achieve continuous positive growth during the year.
In addition, according to the survey results of "China Industrial Internet Platform Market Analysis 2021" released by Another third-party analysis agency IDC (Internet Data Center), 45% of China's industrial Internet platform enterprises have a compound revenue growth rate of 60%-100% between 2018 and 2020, that is, the market as a whole is in a state of rapid growth. But at the same time, in 2020, 65% of key platforms are in a basic breakeven or unprofitable state, and more companies are still exploring how to optimize their business models to achieve profitability.
At present, Baidu Intelligent Cloud is also facing the need to improve the functions of platform products. Problems such as insufficient accumulation in vertical areas and shallow cooperation between ecological partners.
At the end of the day, AI and industry need to understand each other. It is inseparable from the multiple integration of technology, resources and talents. If industry has its own beat, then the blessing of AI and multi-party forces should be the same frequency resonance.
Taking the simplest visual recognition as an example, in the consumer Internet scenario, visual recognition can be accurate to "brush face payment", indicating that the technology is very mature, but in the industrial scene, the recognition of flames does not require how sophisticated algorithms, at this time what is really needed is to identify under the interference factor - just a gust of wind blowing, or there is water on the ground, the size and shape of the flame may be completely different.
Baidu Intelligent Cloud positions itself as a "digital base, intelligent engine, empowering thousands of industries", which means focus and boundaries: Baidu Intelligent Cloud is responsible for providing native scientific and technological power, and cooperation with ecological partners at the application level becomes particularly important. Cooperation has therefore been given its due meaning: one is that it must be open, the other is that AI capabilities are the underlying technology, and the solutions that Baidu Intelligent Cloud really outputs also include the understanding of intelligent transformation and the ecological products that can be driven.
On December 27, Baidu founder Robin Li said in his speech at Baidu Create 2021 (Baidu AI Developer Conference) that in the era of "human-machine symbiosis", China will usher in the golden decade of AI. In the next decade, AI will significantly reduce the technical threshold and provide a technical "big base" for the intelligent transformation of all walks of life.
Baidu chief technology officer Wang Haifeng said that the artificial intelligence platform Baidu brain open capabilities nearly 1400 items. The domestic deep learning platform Flying Propeller brings together 4.06 million developers and serves 157,000 enterprises and institutions, ranking first in the comprehensive market share of China's deep learning platforms.
This summer, Baidu Intelligent Cloud also released the "Open Things" industrial Internet platform, from the four levels of point, line, surface and body, to provide AI+ industrial Internet and intelligent manufacturing overall solutions for industrial enterprises, industrial chains and regional industrial clusters.
An Internet executive told us that at present, there are new requirements for technological self-reliance. It is generally believed that the "card neck" is generally the "card neck" in traditional technology, and when Chinese enterprises encounter the problem of "card neck", it coincides with the opening of the fourth industrial revolution and the arrival of a new technological era. It means that all countries are on the same starting line again.
"The problems we have encountered in the industrial Internet have not been encountered abroad."
To solve the "card neck", there is a new solution is to change lanes, not to overtake in curves, but to change lanes to overtake. If the "stuck neck" problem is solved with a lane change, the cost will be much smaller and the time cycle will be shortened.
This is an opportunity for platform companies, and it also means that you must abandon your illusions and grow downwards.
Four years ago, when Baidu Smart Cloud just decided to start doing industry, there were only 30 people. At that time, Li Shuo himself had no bottom in his heart, and he did not know whether his team could create value. He simply reassured colleagues, "Such a big industry and track, we left an exploration team in this place." ”
Today, in Chongqing's Liangjiang New Area, Suzhou's Changshu, as well as Shenyang, Quanzhou, Guangzhou, from north to south, Baidu's intelligent cloud industry team has measured the foundation of China's economy inch by inch, and the end point has never changed - using technology to make the complex world simple.