AI pork is coming, what exactly is it?
From its founding in 1991 to now, private pig enterprise Jilin Jingqishen Organic Agriculture Co., Ltd. (hereinafter referred to as “Jingqishen”) has raised mountain black pigs for 28 years, supplying products to Walmart Sam's Club, Carrefour, AEON, China Resources Ole, JD 7Fresh and others. After the 40-year-old Jingqishen CEO Sun Yanchun took over the family business, he had new calculations in mind: besides actively introducing mechanized equipment to improve efficiency, could more intelligent means also be introduced to raise pigs and improve management processes?
For example, one difficult problem in pig farming is how to count the exact number of tens of millions of piglets, and how to recognize each pig. Going further, from the perspective of the farm's management process, if the power of machines could be used to prompt employees “at what time node they should appear at what position”, could work efficiency be greatly improved?
“The logic of these technological advances sounds simple, but traditional enterprises like ours cannot do this kind of research and development. So we urgently have this need, and we need some high-tech enterprises and enterprises with resource integration capabilities to do this.” Sun Yanchun said.
1. Pig face recognition has arrived
On June 13, at the Jingqishen breeding base in Fusong, Baishan, Jilin, Li Jialong, deputy general manager of JD Agriculture and Animal Husbandry, introduced that the initial research and development of pig face recognition technology was not part of the current “smart livestock solution”, but for the needs of JD's insurance team.
“This is a long story. We started doing pig face recognition because JD insures pigs. For example, if 100 pigs are raised and 10 of them are insured, then no matter which pig dies, I can go and claim reimbursement, because the insurance company also does not know which 10 of these 100 pigs are the ones I insured, so this problem exists: if any of the other 90-plus pigs die, I can also go and claim reimbursement. JD's insurance team very much hoped to use a method to recognize these 10 pigs, so we began researching pig face recognition technology in 2017.” Li Jialong said: “Starting in September of that year, we often went to pig farms to collect data and learned about the current state of China's livestock industry, so we began doing JD Agriculture and Animal Husbandry.”
China is a major livestock-raising country, with its live pig inventory ranking first in the world, and pork output accounting for about half of the global total, but in terms of labor productivity, feed conversion rate, sow productivity level, and production costs in pig farming, there is still a certain gap from the international advanced level.
This became the background for the establishment of the JD Agriculture and Animal Husbandry team and for multiple internet companies entering agriculture: in the world's largest pig farming and consumption market, use intelligent means to carry out transformation and upgrading, in order to reduce part of the costs in pig farming and improve production efficiency.
In March 2018, the JD Agriculture and Animal Husbandry team began experiments. In December 2018, JD officially invested in Jingqishen and obtained a 9.43% equity stake. Jingqishen currently has 1,600 head in its core herd and 8,200 basic sows. It currently has 7 large-scale breeding pig farms in Jilin Province and 275 cooperative households with inventory, with an annual slaughter of 200,000 commercial pigs, making it a medium-sized pig farming enterprise.
2. How does AI raise pigs?
In February 2019, after the Spring Festival, several developers from the JD Agriculture and Animal Husbandry team entered Jingqishen's pig farm in Fusong, Baishan City, Jilin. According to the JD Agriculture and Animal Husbandry team, in more than 100 mountain black pig barns across 2 breeding parks in Baishan, Jilin, Jingqishen deployed JD Agriculture and Animal Husbandry's smart breeding solution based on AI, IoT and SaaS technologies, including Shennong Brain, Shennong IoT devices, and related systems and hardware equipment of the Shennong system.
On June 13, the author saw in Jingqishen's mountain black pig barns that the barns had already been equipped with breeding inspection robots, feeding robots, agricultural-grade cameras, retractable semi-limiting pig pens and other facilities. Through the above equipment, the application of image recognition and artificial intelligence technology is flexing its muscles in the agricultural field.
The inspection robot is installed on the ceiling of the pig barn and moves back and forth along the inspection vehicle track at the top, using cameras to inspect activity inside the barn.
According to the JD Agriculture and Animal Husbandry team, taking pig counting as an example, ordinary breeding enterprises conduct inventory counts at the end of each month. Traditional naked-eye counting of pigs is very inefficient and inaccurate, and it often takes 3 to 4 days to complete the inventory count. As for weighing pigs, traditional breeding also only weighs them once when they are sent to slaughter, and before that one can only estimate based on experience.
The inspection robot uses a dedicated agricultural camera and can know the number of pigs with a “glance” at the pig pen, and estimate the weight of each pig. Behind this is the support of AI algorithms.
The camera in front of the feeding pen is part of the intelligent feeding robot and can perform “pig face recognition” on pigs: each feeding compartment accommodates only one pig, and when the pig walks in, its image information can be captured by the camera.
In the traditional mode of feeding, pig feeding relies on manual feeding of 10 to 15 pigs together, and each pig can only compete by strength to eat. As a result, pigs in the same pen often differ in weight when sent to slaughter.
Through “pig face recognition” technology, the “identity information” of each pig can be identified, thereby retrieving its weight and growth cycle information, enabling more precise automatic feeding and minimizing the weight differences among pigs in the same pen when sent to slaughter.
It is worth mentioning that the outside world also has different voices on the practicality of “pig face recognition” technology. Previously, other methods for solving identity recognition and quantity calculation already existed on the market, including putting electronic smart “earrings” on live pigs, or making tattoo codes on pigs' backs and importing them into the system through cameras.
However, Li Jialong believes that the above methods still generate a large amount of cost during installation and also have a certain impact on the pigs themselves, while “pig face recognition” can more effectively avoid the occurrence of the above situations. Li Jialong introduced that when “pig face recognition” technology is specifically applied, it requires about 30 to 50 days of localized deployment.
In addition to the above device-side installations, the core of JD Agriculture and Animal Husbandry's technical solution is the “Shennong Brain” developed using artificial intelligence technology. Simply put, the “Shennong Brain” stores professional pig farming knowledge and can independently make judgments in various situations.
For example, if a pig in the barn suddenly has a weak squeal and also coughs occasionally, Shennong Brain will first detect the abnormality and require the camera to observe the pig's body temperature, call up the pig's past body temperature, feeding records, and activity amount in the database for comparison. If the squeal, body temperature, and movement are all abnormal, then Shennong Brain can make a preliminary judgment that the pig is very likely sick, and next, Shennong Brain will send an alert to the breeder.
3. Doing the math on AI pig farming
However, installing these cameras and robot devices is still a relatively large investment for farmers. Jingqishen's breeding volume is medium-sized, and it can achieve a certain scale effect in saving costs, but for more small and individual farmers, before using this intelligent equipment system, they inevitably need to calculate the input-output ratio well.
The Jingqishen team said: “At present, after we invested in the AI breeding system, taking Jingqishen Beishan Pig Farm and Xingshen Pig Farm as examples, we expect an average increase of 92.5 yuan in cost per pig (total investment, calculated based on 50,000 pigs slaughtered per year for 8 years, totaling 400,000 pigs), but in terms of feed, 30 to 60 yuan can be saved per pig; labor cost input can save 8 yuan per pig; in improving the survival rate, an average of 15 yuan can be saved per pig; in precisely controlling slaughter weight, 5 to 8 yuan can be saved; totaling 91 yuan saved. From the perspective of direct input and output, we are basically breaking even.”
However, as the duration of use increases, the average cost may further decrease.
In terms of feed costs, the Jingqishen team stated in written materials that “through tests in the experimental group over nearly a year in the earlier period, we found that we can achieve at least about a 10% increase in feed conversion efficiency for pigs. For us, with an annual slaughter of 200,000 mountain black pigs, this will be a huge benefit, and preliminary estimates can reach an annual savings level of 12 million yuan.”
More importantly, intelligent technology also brings changes in management thinking and processes. “The key is to reduce employees' blind work and useless work during work, and at the same time also manage the work process and results through the system, and analyze and do a good job in managing employee work performance.”
Sun Yanchun gave an example: in the past, many rules and regulations were designated in the breeding farm, but their implementation relied on employees' sense of responsibility, and it was difficult to measure employee performance. Now, relying on intelligent equipment and SaaS enterprise operating systems, management can be more refined. “For example, if there is a place in the pig farm with a lot of manure spots, a robot inspection sees the manure spots, publishes information prompting an employee to clean it up. If no one goes in the first 10 minutes, the employee himself is prompted; if no one goes in the second 10 minutes, the supervisor is prompted; if no one goes in the third 10 minutes, it is reported to the factory director. After both targeting and compliance have been changed, the quality and efficiency of employees' work can truly revolve around the core value of the enterprise.”
Sun Yanchun introduced that through intelligent management equipment and information prompts, one farm director can manage more pig herds. “Take our farm. In the past, there was one farm director, two deputy farm directors, and seven or eight technicians, such as those managing the farrowing stage, the mating and gestation stage, and the nursery stage. Now after our optimization, only one principal farm director, one deputy farm director, and three or four technicians are needed to complete this work.”
Jingqishen also stated that “apart from the most direct comparison of input and output above, the environmental protection value, breeding value and management value that the (intelligent breeding) system brings us, as well as the basic depreciation of the system's software and hardware, are values that have not entered the scope of comparison. Overall, this is a very cost-effective project.”
One area with huge potential is that the intelligent system's collection and monitoring of breeding information may be able to leverage the financial industry to increase support for agriculture.
4. JD Agriculture and Animal Husbandry smart breeding solution
The Jingqishen team affirmed the role of the above “Shennong Brain” at the practical operation level. “When breeding personnel observe the health of the pig herd, besides being limited by technical level and means, a more important point is that they cannot spend enormous energy and manpower to observe the health of every pig. This leads to the awkward situation where once there is disease, it affects a large area, and it is already the terminal stage of the epidemic. With an intelligent system, the system will monitor each pig's behavior, movement, body temperature, feed intake, and even voiceprint information one by one, and after calculation determine abnormal individuals and mark them to guide staff follow-up and treatment.”
Regarding the barn environment, Shennong Brain will also conduct unified management and intelligent analysis of the collected pig farm environmental parameters, and then adjust them through intelligent fans, intelligent oxygenators, intelligent humidity regulators and other intelligent equipment, ensuring that the temperature, humidity and air of the pig farm are maintained in the best state suitable for the healthy growth of live pigs.
The above systems and equipment actually process the growth situation of each pig into information. According to JD Agriculture and Animal Husbandry's plan, from a pig's birth, growth, quarantine, slaughter, processing, and transportation to sales, information at each stage is intended to be traceable.
“Jingqishen will synchronously transmit the big data of JD and Jingqishen intelligent breeding to the Jilin Provincial Animal Husbandry Administration through the internet, and merge biosafety data and production traceability data on the same track into the (Jilin) provincial bureau's epidemic prevention management platform and the ‘Gong e Gong’ product traceability platform.” the Jingqishen team said.
Not only daily breeding and management. In Sun Yanchun's plan, in terms of breeding, the use of intelligent means may also greatly improve breeding efficiency.
5. Visualizing various types of information from the breeding farm
Sun Yanchun mentioned that in the past banks were quite cautious about loans to agricultural enterprises. “We had to use a lot of collateral, which was very painful. Our enterprise's pig barns were not recognized by them, nor were the pigs. We used our hard assets, the factory's real estate, land certificates, and equity as collateral.”
“Now there is no problem. The (inspection robot) running on this (referring to the breeding farm) runs one lap, and how many pigs are inside can be counted very clearly at any time and place. The bank can see it very clearly, and now they are reassured.” Sun Yanchun said, “These important pieces of information, in the future, we will proactively connect to the government competent department and the financial institutions and insurance companies that provide us with credit services, so that the industry competent department and business service departments can sit in their offices as if they were on the front line, know the inventory status and health status of the pig farm, and let everyone carry out management and service work with peace of mind.
According to JD Agriculture and Animal Husbandry, more and more breeding enterprises have begun to come into contact with smart solutions from internet companies. Not only pig farming. Li Jialong introduced that JD Agriculture and Animal Husbandry's smart breeding system will also play a role in cattle farming, chicken farming, aquaculture, planting and other aspects.
“AI Fresh Meat Shop” appears
Before this year's “6·18”, a topic about the “AI Fresh Meat Shop” made it onto the Weibo hot search list, sparking heated discussion among netizens. This pork shop named “AI Fresh Meat Shop” is located in the JD 7Fresh supermarket store at Dazu Plaza in Beijing. The store is clean and tidy and full of technological flavor, very eye-catching, and looks very high-end. The “AI Fresh Meat Shop”, jointly launched online and offline across all channels by JD Agriculture and Animal Husbandry under JD Digits and Jilin Jingqishen Organic Agriculture Co., Ltd., stems from the fact that the pork sold in the fresh meat shop is pork produced using artificial intelligence technology.
AI pig farming should be given attention
In 1956, “artificial intelligence” was proposed. In 1997, the “Deep Blue” computer defeated chess grandmaster Kasparov. In 2017, “artificial intelligence” was selected as one of the “Top Ten Chinese Media Buzzwords of the Year”, with application areas covering computer science, finance and trade, medicine and health, industrial transportation, telecommunications, online services, legal consulting, scientific discovery, music and games, planting and breeding, and other aspects. In 2019, the Second Session of the 13th National People's Congress included legislative items closely related to artificial intelligence in its legislative plan.
In February 2018, Alibaba cooperated with Huaxi Hope Tequ and Sichuan Dekang Agriculture and Animal Husbandry, aiming to complete the development and training of smart pig farms through joint efforts and information sharing by all parties, and promote the development of China's pig farming industry toward intelligence; in June, Alibaba Cloud launched the “ET Agricultural Brain”, applied to pig farming, apple and melon planting, with functions such as digital file generation, full life cycle management, intelligent farming analysis, and full-link traceability.
In August 2018, JD Cloud, Nxin Internet and Tianzhao Pig Industry established the Smart Pig Farming Alliance; in November, JD established an agriculture and animal husbandry company to enter the pig farming industry, planning to cooperate with China Agricultural University, the Chinese Academy of Agricultural Sciences and others to research, develop and launch a smart breeding solution, with the main goals of reducing labor costs for medium and large pig farms by more than 30%, saving 8% to 10% of feed, and shortening the slaughter time by about 1 week.
At present, “African swine fever” is raging globally, and pork has always occupied a prominent and important position in the meat consumption of Chinese residents. People are full of expectations and attention for AI pig farming.
According to analysis, a mature AI pig farming model, by adopting advanced information, communication and IoT technologies to achieve highly automated control of the entire production process on pig farms, can greatly save labor, reduce feed input, improve growth and reproductive performance, increase production efficiency, achieve real-time monitoring and effective control of diseases, make the entire pork production process traceable, and ensure food quality and safety. Compared with traditional breeding, it indeed has obvious advantages.
The pig farming industry itself is an industry with large investment and a long capital recovery cycle, while AI hardware equipment is expensive and requires huge investment, and its stability and compatibility have not yet been well integrated and verified. The normal operation and maintenance of facilities and equipment place high knowledge requirements on professional and technical personnel, and existing equipment and systems have not yet formed a mature model and reliable capability to precisely guide the entire production process of pig farms of different scales and types.
Science and technology are the primary productive forces, and artificial intelligence is advancing by leaps and bounds. We should indeed be full of expectations for AI pig farming and treat it positively. In light of the current situation, for the development of AI pig farming, the government should reasonably arrange special funds to encourage scientific and technological innovation, recognition of achievements, technology integration and demonstration promotion, and also formulate policies and measures conducive to scientific and technological innovation, guide more social resources to invest, and encourage the formation of institutional mechanisms for concentrating firepower on joint tackling of key problems and concentrated breakthroughs, avoiding relevant entities fighting separately and wasting social resources. Encourage capable industry-leading enterprises such as Alibaba and JD to join with scientific research institutes, and take on the role of explorers, innovators, practitioners and promoters of AI pig farming. For the many small and medium-sized pig farms, it is better to pay timely attention to scientific and technological progress and equipment innovation in related fields and share the fruits of scientific and technological development with an open mindset. For any emerging technology, not only should we pay attention to whether it is technically feasible, but we should also consider whether it is economically feasible and optimal.