On April 1 last year, Tencent released an AI Goose Factory project, claiming to use artificial intelligence technology such as "goose face recognition" to build AI agro-ecology, which is obviously a April Fool's Day joke.
However, in the year of looking back in 2019, it was not only a simple joke.
Following Alibaba’s announcement in February last year that the ET brain will be used to achieve AI pig raising, JD.com (formerly JD Finance) also recently launched an intelligent farming program based on AI technology such as “pig face recognition”. Netease may not be able to Enjoy the title of "Pig Factory".
Although Tencent’s AI geese have not been realized, they have planted cucumbers with AI. Jingdong, who just joined the pig-raising army, has fallen into more than 10,000 square meters.Plant factoryIt is said to increase production to 3 million tons per year, which is several times the efficiency of traditional vegetable farmers.
Internet giants have entered the agricultural industry, and there must be huge commercial value behind them. What new changes will these giants bring to traditional agriculture?
What did the Internet company do in the field?
Among many Internet companies, NetEase is one of the first companies involved in the breeding industry. Moreover, the self-operated farm farming model adopted by NetEase Weiyang penetrates into the entire chain from farming to sales.
▲ Ding Lei (after) and Netease Weiyang pig
Netease Weiyang advocates humanized farming. Here, the black pig sleeps in the apartment, listens to music, picks up the toilet, drinks deep groundwater, and eats the liquid pig food. It is said that people can also eat it directly, so many people lament that "people are not as good as pigs". .
However, Ali and Jingdong, two Internet companies that entered the pig industry last year, are very different from Netease. They have not used nutrition farms to help the traditional aquaculture enterprises to rebuild their culture. mode.
Although NetEase also uses a fully automatic liquid grain feeding system in pig farms, the two companies want to do more with the technology to transform the pig industry.
Ali uses AI for pregnancy testing
In February last year, Alibaba Cloud officially announced the cooperation between Sichuan Special Drive Group and Dekang Group, and realized artificial intelligence raising pigs through ET brain to improve the survival rate and calving rate of pigs. The project invested hundreds of millions.
So how is the so-called AI pig raising?
Here we have to first understand the two most important indicators for the pig farm ——PSY and MSY, in simple terms, how many ticks per sow can be produced each year, and how many piglets can eventually grow into fat pigs. At present, the average PSY of pig farms in China is about 15 and that of the United States is 25, which means that each pig in China has 10 fewer pigs per year than the United States.
China currently has about 700 million pigs per year, and the cost of raising a pig is twice that of the United States. The amount of money earned in it is considerable, so the AI pig raises the amount of sows. And let more piglets grow up healthily.
Ali uses a system called ET Agricultural Brain, which involves artificial image technology such as video image analysis, video image analysis, face recognition, speech recognition, and logistics algorithms.
Aliyun developed a "pregnancy diagnosis algorithm" to determine whether a sow is pregnant. A number of automatic patrol cameras in the pig farm collect data on the sow's sleeping position, standing position, eating, etc., and then the AI analyzes whether the sow is successfully matched. If the sow is not pregnant, the system will remind the staff. Artificial insemination is carried out to increase the amount of sows born.
After the birth of the piglets, in order to make them grow up healthily, the ET agricultural brain monitors the health of each pig through speech recognition technology and infrared temperature measurement technology, because the pigs do not sound in different states such as feeding, sleeping and getting sick. Similarly, by analyzing data such as cough, snoring, and body temperature in pigs, an early warning can be issued in the event of an abnormality.
▲ Image from:Ali Cloud
For example, when the piglet is pressed by the sow, the system can be judged in time by the call of the piglet, and the breeder is called to deal with it in time to improve the survival rate of the piglet.
In addition, each pig has a dedicated identity earring that records their weight, eating and exercise intensity, frequency and trajectory. If any pig does not meet the standard but is not sick and pregnant, the breeder will drive the pigs to When exercising outdoors, the ET Agricultural Brain will record how many steps each pig has taken in the same way as the WeChat campaign.
Ali even used the number of kilometers as a new standard for judging the quality of pork. The former Alibaba Cloud President Hu Xiaoming said at the Yunqi Conference in June:
In the future, what we want is a good pig that ran 200 kilometers, instead of a 200-pound pig.
Last year, ET Agricultural Brain has been tested in a farm with 3,000 pigs in Sichuan. It has successfully produced an average of 3 sows per litter, and the death rate of piglets has decreased by 3%. Next year in Sichuan Special Drive Group and Germany The AI system will be installed on the 50 farms of the Kang Group above the size of the sow.
Jingdong's "pig face recognition"
Ali has an ET agricultural brain, and Jingdong launched the "Shen Nong Brain" at the end of the year. This is a smart farming technology developed by Jingdong Agriculture and Animal Husbandry in collaboration with China Agricultural University and the Chinese Academy of Agricultural Sciences.
According to Jingdong, the entire intelligent breeding program includes three modules: Shennong Brain (AI), Shennong Internet of Things (IoT), and Shennong System (SaaS), which can realize "intelligence of pigs."
Among them, "pig face recognition" is the most concerned, and the farm can use this technology to observe and record the weight, growth and health of each pig. Although face recognition has become popular in smartphones, it is much more difficult to use on animals.
Researchers at Michigan State University have developed a set of primates.Facial recognition system PrimNetHowever, because of the large differences in animal hair and eye color characteristics, the research team had to manually mark the animal's eyes and mouth to assist in system identification.
It is more difficult to accurately identify the face of a pig than animals such as monkeys and cattle and sheep. Jingdong Digital Branch official account is inKnowing the answerAmong the three technical difficulties mentioned:
However, "pig face recognition" is not a patent of JD. In March last year, Chen Yaosheng, chief scientist of the national hog industry technology system, officially released "Pig face recognitionIn the technology, the sow recognition rate was 98% in the experimental field and the hog recognition rate was 85%.
The farm can use the “brush face” to identify different feeds according to the state of each pig during the breeding process. By the end of March, there are already168,821 large pigs in 1692 pig farmsParticipated in the pig face recognition test.
Jingdong's intelligent farming technology not only uses the "pig face recognition" to achieve precise configuration of pig feed, but also the temperature and humidity in the pig house are automatically adjusted by the system. At the Jingdong Digital Technology Global Explorer Conference, Li Defa, academician of the Chinese Academy of Engineering Indicates:
In the future, we will start with pigs, realize no breeding, wireless, no interference, no contact in the breeding process, and realize the real Internet of Everything.
According to JD's calculation, the labor cost of this system is reduced by 30%~50%, the feed use is reduced by 8%~10%, and the slaughter time is shortened by an average of 5~8 days. If all the farms in China use this system, the annual cost will be reduced by 50 billion yuan.
Tencent and Ali's sweet melon
Last month, the AI of the AI Lab team of Tencent's artificial intelligence laboratory defeated humans twice. One was that AI "Enlightenment" defeated the human team composed of former professional KPL players in "Glory of the King", and another victory was It is in the AI Cucumber Competition.
In the AI Crop Development and Simulation Business Challenge initiated by Wageningen University in the Netherlands, the iGrow team consisting of Tencent and the Chinese Academy of Agricultural Sciences achieved second place.
▲ Image from:ITQB
This is a cucumber competition for the artificial intelligence team. Each AI team uses machine algorithms to remotely control the cultivation of cucumbers, except for the last process of picking cucumbers.
According to Tencent, watering, ventilation, lighting, fertilization and topping decisions are all collected automatically by artificial intelligence AI, through deep learning and calculations, judgments and decisions, and then drive the equipment components in the greenhouse.
▲ Image from: Tencent News
In addition to cucumber yields, AI strategy and sustainability indicators, the iGrow team harvested 3,496 kg of cucumbers in four months, ranking second.
The pressure to win Tencent is from Microsoft.Sonoma teamIt is the only team that has planted more than 50 kilograms of cucumber per square meter and has a net profit that exceeds the human team.
But AI planting is currently not completely replacing humans, and not all greenhouses are suitable for this model, nor are all crops like cucumbers with a large amount of data that can be used to train AI.
However, Ali's ET agricultural brain has tried to use AI for fruit planting. The agricultural giant Haisheng Group's 10,000-acre apple orchard in Shaanxi has used this system for management. Just scan the QR code on the fruit tree. Historical watering, fertilization, and pesticide application.
In addition to apples, the current application of ET agricultural brain has been extended to citrus, berry, carrot, pear, cherry, kiwi and other categories, which can help fruit farmers save more than 200 yuan per mu.
In the future, AI planting may really be as convenient as playing a happy farm, further liberating manpower.
Yuan Longping, who wants to be an Internet age, still has a long way to go.
In fact, in the current agricultural market, there is no shortage of high-tech figures. From automatic spraying of agricultural drones to unmanned tractors, the degree of automation of agricultural production is increasing, and AI is an important driving force behind bridge.
China's Internet giants are eager to experiment in the agricultural sector, and a large number of AI agricultural startups have emerged in foreign countries.
Blue River, an AI startup that was acquired in 2017 for $305 million, uses machine vision to identify crops and weeds for smart spray.
Israel's AI startup Prospera helps farmers analyze agricultural data collected through computer vision and artificial intelligence.
Berlin's agrotech startup PEAT has developed a deep learning application called Plantix that uses algorithms to identify nutrients and crop pests that are lacking in the soil.
Currently on a global scale, agriculture is a value3 trillionIn the US dollar industry, for Internet companies with e-commerce genes such as Ali and JD.com, using technology to transform agriculture can help traditional agriculture improve production efficiency, and it is also beneficial to cut into agricultural products market through its own e-commerce platform and logistics system, such as Ali. "Rural Taobao."
It is good for both the country and the people, and it is of great commercial value. In a speech, Ma Yun talked about the vision of agricultural innovation:
Mr. Yuan Longping has achieved a thousand pounds of production per mu. Our Internet technology must strive to achieve a thousand dollars in per mu.
However, if it is not easy to become "Yuan Longping in the Internet age", even in the case of the pig industry, even if AI can significantly improve the efficiency of breeding, the farms in the rural areas of the 4th and 5th tiers often do not have high speeds and the infrastructure is not. Perfection is a big obstacle.
In addition, the current cooperation with Ali Jingdong is basically a large-scale farm, but in the whole market, there are actually more retail investors. It is very difficult for one family to persuade these retail investors to access the AI system.
Zhang Tengfei, the CTO of Ruiye Technology, who also works as an AI pig system, mentioned in a speech:
China's largest aquaculture company is Wen's in Guangdong, but it only accounts for 3% of China's culture. The top eight in the pig industry, more than half of which are our customers, but their farming volume only accounts for 7% of the industry. More than 50% of pork in China is raised by small and medium-sized retail investors.
However, agriculture is also an important criterion for verifying whether artificial intelligence leads the fourth industrial revolution. In 1995, the academic community proposed a"General Purpose Technology"The concept of defining revolutionary technologies that have a huge, far-reaching and widespread impact on the human economy and society, with several main criteria:
At present, there is still controversy in the industry about whether AI belongs to "general purpose technology". Erik Brynjolfsson, author of "Second Machine Revolution", believes that AI belongs to "general purpose technology", but it must wait until the corresponding infrastructure and supporting components are perfect and the cost is low enough. After the explosion will come.
If the AI technology of these Internet companies can be widely applied in the agricultural field in the future, this view has basically been verified.