Industry News

AI breaks the boundaries of food research and development in the laboratory, accurately schedules ingredient inventory in the supply chain, and restores the "master fire" in restaurant kitchens... Currently, the food industry is undergoing a full chain digital upgrade. This intelligent transformation from research and development to consumption not only reduces costs and increases efficiency for food enterprises, but also adds more diverse taste choices for ordinary people. What substantial changes has the food industry undergone under the accelerated penetration of artificial intelligence? How to better meet the personalized needs of consumers for nutritious meals? What new imaginations will AI bring to the development of the industry in the future?

 

This year's Government Work Report proposes for the first time to "create a new form of intelligent economy". Currently, artificial intelligence is widely and deeply integrated with various industries and fields, bringing new imaginative space for industrial development. The food industry is a traditional advantageous industry and an important livelihood industry in China. With the all-round empowerment of artificial intelligence in various industries, the food industry is actively embracing new technologies and promoting the deep application of cutting-edge technologies such as artificial intelligence models in research and development, pilot testing, production, and management.

 

The support from the policy side injects momentum into the digital transformation of the food industry. The Implementation Plan for Digital Transformation of the Food Industry, issued by seven departments including the Ministry of Industry and Information Technology in June 2025, proposes to encourage innovative application of artificial intelligence technology and enhance the supply of artificial intelligence large-scale model technology products in the segmented food industry; Supporting cooperation between enterprises, research institutions, and platform enterprises, utilizing data twins, artificial intelligence big models, etc., to achieve intelligent perception, lean manufacturing, intelligent detection, and precise delivery, and to build a new data-driven and collaborative operation and management model for production, sales, and use.

 

What kind of intelligent transformation has occurred in the food industry when it encounters artificial intelligence, from research laboratories, enterprise production lines to consumer dining tables? The reporter conducted an in-depth interview on this matter.

 

A brand new landscape at the forefront of scientific research

 

In traditional food research processes, data acquisition often involves tedious sample pre-processing, repetitive instrument operations, and extensive manual intervention. But in a special laboratory at the Institute of Agricultural Products Processing, Chinese Academy of Agricultural Sciences, a completely different scientific research landscape is unfolding.

 

After waking up in the morning, researchers send remote commands to robots. When they enter the laboratory after breakfast, the data has already been measured. This is not a distant imagination, but a technological scenario that has already been realized, "said Ma Peihua, a researcher at the Institute of Agricultural Products Processing, Chinese Academy of Agricultural Sciences.

 

Ma Peihua introduced that his team's "fully automated food nutrition analysis laboratory" is an integrated laboratory with an automation level of L3, and the laboratory's automation capabilities are still being continuously upgraded. Taking the pre-treatment process of agricultural product samples as an example, robots can directly complete the entire process from sampling with a drill bit to homogenization, which previously often required manual use of household mixers. Robot operation is not only more efficient, but also reduces the workload of researchers.

 

Ma Peihua's team is currently conducting large-scale intelligent testing based on automated laboratories, collecting large amounts of data to support the construction of large-scale food models.

 

The development of large-scale food models is a systematic project. Ma Peihua believes that the development process should focus on the accuracy of large models. "For example, when we want to take a photo to identify the calorie or polyphenol content of food, we naturally hope that it can be fast and accurate, but 'accuracy' is the first priority, even more important than 'speed'.

 

Honesty is the technical bottom line of food modeling. The latest research has found that as the accuracy of large models improves, AI has become adept at "deception" - it knows the correct answer but outputs incorrect information. Ma Peihua believes that the nutrition and safety of food are crucial for human health, and the research on food models should pay special attention to their honesty, which is an indispensable defense line.

 

We are using AI to rediscover science. ”Ma Peihua said, "From phenomenon recognition to mechanism discovery, to risk warning and functional design, AI is driving food science towards an interpretable, predictable, and creative future. ”

 

Ma Peihua said that the team will develop various innovative applications based on the big food model to meet the industry's multiple needs for personalized nutrition, food safety, etc.

 

Not only in research laboratories, but also in the R&D end of enterprises, AI is playing an increasingly important role.

 

The application of AI is very extensive in market research and user demand collection. In the past, conducting market research required field research, family interviews, and phone interviews, which took at least one or two months. Nowadays, using AI research can quickly and efficiently sort out consumer needs, and research results can be obtained in just one or two days, greatly accelerating the pace of enterprise R&D innovation. ”Zhejiang Baby Greed Food Technology Co., Ltd. Chairman Guo Baoping said.

 

Yu Qinghua, Senior Research Director of Difule Biotechnology (Shanghai) Co., Ltd., pointed out that artificial intelligence is profoundly changing traditional research methods from the bottom and reshaping the research and development model of healthy food.

 

Yu Qinghua introduced that currently, Difule is focusing on implementing AI applications in two directions. One is AI analysis and application based on human symbiotic microecological data. Difule has independently developed a "Gastrointestinal Microecological Health Assessment and Personalized Solution", which combines gut microbiota gene testing and artificial intelligence analysis to help researchers, clinical doctors, and consumers identify potential health risks earlier and more accurately. The second is the intelligent upgrade of micro ecological research and clinical research design. By analyzing massive existing data through AI, relevant research can be conducted more efficiently and accurately.

 

Yu Qinghua believes that artificial intelligence is helping researchers gain a deeper understanding of the complex "micro ecosystem" of the human body. "In the future, we hope to further integrate artificial intelligence, big data, and micro ecosystem research, gradually shifting health management from experience driven to data-driven and science driven.

 

In addition to mining clues from data like an analyst, AI can also become an engineer, making production lines more flexible and intelligent, and making quality control easier and more efficient.

 

Zhao Yunjiao, Director of Application Technology at Weikang Probiotics, introduced that in terms of research and development, Weikang uses a microbiome database and large-scale strain resources, combined with AI algorithms for strain screening and functional prediction. Through data modeling, it shortens the strain screening cycle and improves the accuracy of functional positioning. This data-driven strain mining model significantly improves research and development efficiency.

 

On the process side, we have built an intelligent fermentation control system that monitors data in real-time, optimizes algorithms, dynamically adjusts fermentation parameters, improves fermentation stability and batch consistency. At the same time, we use data modeling to predict risks during the process amplification, reduce trial and error costs, and achieve efficient conversion from the laboratory to the production line. Zhao Yunjiao said that in terms of quality control, the company uses an intelligent detection and data traceability system to achieve traceability management throughout the production process, ensuring that the activity, purity, and stability of each batch of bacterial powder products meet standard requirements.

 

Intelligent assistant in the processing stage

 

Accompanied by the sound of machine operation, dried locust flowers mixed with impurities such as stones and wooden poles move forward along the conveyor belt. At the other end of the conveyor belt, the industrial camera at the head of the sorting robot blinks rapidly, taking dozens of photos in one second and accurately locating the coordinates of impurities. Subsequently, the high-pressure nozzle of the robot will blow impurities and unqualified materials into the waste bin, while qualified dried locust flowers will enter the next processing stage along the conveyor belt.

 

This is the intelligent sorting scene that the reporter saw in the testing workshop of Beijing Minyong Digital New Technology Co., Ltd.

 

Peanuts, dehydrated vegetables, tea, goji berries... The production and processing of many foods rely on the selection and sorting of ingredients. How to complete this step with high quality and efficiency?

 

The application of artificial intelligence technology in sorting robots provides a new option for food enterprises. Ding Gangtao, founder of Beijing Minyong Digital New Technology Co., Ltd., introduced that the traditional manual sorting method is time-consuming and labor-intensive. Previously, the company had also replaced some manpower with optoelectronic equipment, but some equipment was not selected cleanly enough, often making it difficult to meet customers' requirements for high quality.

 

The introduction of AI visual inspection technology in sorting robots has made up for the shortcomings of manual sorting and optoelectronic equipment. After the fabric system spreads the material onto the conveyor belt, the sorting robot will first take an image, analyze it, and then issue instructions. ”Ding Gangtao said.

 

Traditional manual food sorting emphasizes "quick eyes and quick hands", while an intelligent sorting robot must have "eyes", "brain", and "hands" to perform the same job. Among them, high-speed industrial cameras are the "eyes" of sorting robots, with high frame rates, stable picking standards, and the ability to work continuously for 24 hours; The "brain" of the robot is embedded with AI vision detection algorithms based on deep learning. By summarizing the intrinsic features of various foreign object images, the machine can "recognize" such substances and accurately identify various foreign objects through images in subsequent work; The "high-speed spray valve+high-pressure air" form the "hand" of the robot. After impurities are identified and located, the robot will accurately remove them with its "hand".

 

The large model cloud platform that is matched with the sorting robot is like a "model supermarket" open to customers. The reporter saw on site that when searching for the keyword "goji berry" in the platform search bar, the system would return 7 entries, including model names such as "goji berry" and "black goji berry", with corresponding labels indicating defect types such as "blackening", "whitening", "crushing", and "withering". Similarly, searching for "Sichuan peppercorns" on the platform will bring up 5 items, including "red Sichuan peppercorns", "green Sichuan peppercorns", etc. The corresponding model labels display screening options such as "blackened", "small stick and thin stem", "stone", "open petal", etc. These labels represent both the sortable food categories and the trained impurity recognition models.

 

Ding Gangtao stated that the core value of this cloud platform is to enable customers to independently define sorting criteria. At present, thousands of model files have been accumulated on the platform, covering multiple fields such as food and traditional Chinese medicine decoction pieces. Customers can directly query, download and use them without the need for tedious analysis and model training, greatly reducing the threshold for use.

 

In what directions will AI visual sorting devices continue to evolve in the future? Regarding this, Ding Gangtao admitted that there is a certain "thinking" time difference in the execution process of current AI algorithms - because the "brain" needs computation and reasoning, the response speed is not yet fully comparable to the instantaneous response of optoelectronic devices. However, in the future, this aspect will definitely continue to improve and move forward.

 

Technology implementation in catering scenes

 

The implementation chain of AI technology is constantly extending from scientific research and innovation to processing and manufacturing, and then to catering services. In recent years, various "smart restaurants" and "AI canteens" have emerged one after another, and various types and multi-purpose embodied intelligent robots have been integrated into the catering scene to achieve autonomous collaborative operations throughout the entire process from ordering, production to delivery. The catering industry, which was once supported by stoves and craftsmanship, now not only retains the familiar "fireworks atmosphere", but also adds a bit of hardcore "technological style".

 

As noon approached, when the reporter arrived at an "AI canteen" in Dongcheng District, Beijing, a long queue had already formed. The interior space of this "AI canteen" is not very large, with several neatly arranged dining tables for four people in the lobby area, and a relatively spacious area in the inner room, equipped with two large round tables that can accommodate ten people.

 

Consumers who have already pre stored their value can come to the pickup area and swipe their meal card or undergo facial recognition verification. After passing the verification, they can receive their meal plates and pick up their meals. The reporter saw that neatly arranged stainless steel containers held various freshly baked dishes, and there was a black sensing area on the countertop in front of the container, marked with "put the plate first, then take the clip". The reporter followed the staff's instructions and placed the plate in the designated area. They used a clip to pick up the dishes and placed them on the plate. The countertop screen immediately displayed information such as the name, weight, price, and nutritional content of the dishes.

 

During self-service meal collection, consumers can also clearly see the interior of the kitchen through the "bright kitchen and stove" - except for the staff responsible for delivering dishes to the pick-up area, there are no professional chefs on site, only a few smart metal stir fry machines operating in an orderly manner. The various dishes displayed on the dining table are cooked by these stir fry machines.

 

In addition to catering robots that provide services to consumers in front of the stage, artificial intelligence also plays an irreplaceable role in the "behind the scenes" links of the entire catering industry chain.

 

The relevant person in charge of Xiabu Xiabu Group introduced that from upstream supply chain, logistics transportation to terminal restaurant services, the group relies on artificial intelligence and digital technology to build a sustainable circular operation system covering the entire industry chain. In the supply chain management process, AI first enters the "strict selection" checkpoint, using digital tools to accurately evaluate and screen suppliers, and building a solid food safety defense line from the source. At the same time, in terms of lamb supply, the group has introduced the Smart Wheel Grazing 2.0 system, which utilizes remote sensing monitoring technology to achieve dynamic planning of grazing paths on its own pastures, ensuring the stability of lamb quality.

 

In the inventory and logistics process, Xiabu Xiabu deepens the construction of the supply chain middle platform, upgrades the supply chain management system, introduces the supplier management inventory model, optimizes procurement strategies, effectively accelerates inventory turnover, and improves warehouse utilization and accuracy. Supported by intelligent scheduling and route optimization, Xiabu Xiabu has integrated the entire process of procurement, warehousing, and transportation, achieving full link data connectivity and visual management.

 

For ordinary consumers, artificial intelligence brings not only a richer dining experience, but also more diverse nutritional choices. Currently, artificial intelligence technology is striving towards meeting the precise nutritional needs of individuals. Xu Haiquan, a researcher at the Institute of Food and Nutrition Development of the Ministry of Agriculture and Rural Affairs, stated that from a production perspective, artificial intelligence technology can customize production according to the nutritional needs of different groups, achieving precise adaptation of the nutritional characteristics of ingredients to the needs of consumer groups. "In the future, personalized nutritional meal customization can also be better carried out, and even automatic matching and ordering of household food according to dietary recommendations can be achieved.

 

The front-end of scientific research is assisted by AI, the production line is empowered by AI, and the daily dining table is accompanied by AI. Artificial intelligence is accelerating its integration into various aspects of the food industry. However, during the investigation, the reporter found that there is a significant differentiation in the industry's understanding and application of artificial intelligence. Some laboratories are equipped with cutting-edge technologies such as robot operation and digital twins, but there are also many laboratories that still rely on researchers to manually complete pre-processing and basic experiments. In the sorting process, some companies have introduced intelligent sorting robots, while more companies still use traditional photoelectric equipment combined with manual labor. Small factories and catering entrepreneurs rely mainly on manual labor to complete the sorting of raw materials and ingredients.

 

Enterprises of different scales also have different attitudes towards AI. The head of a small and medium-sized enterprise bluntly stated, "We are small in size and currently cannot reach the threshold of artificial intelligence." Some companies equate artificial intelligence with concrete robots and lack awareness of the application of digital technology. Some companies hold a reserved attitude towards the introduction of AI technology due to their product positioning as "pure natural" or "ancient manufacturing".

 

This difference in cognition and technology reflects the reality that AI faces in the process of penetrating the food industry - the implementation of technology not only depends on its own feasibility, but also on the scale, cost bearing capacity, and conceptual identity of enterprises. Therefore, there is no unified intelligent "template" for the food industry, and different enterprises have chosen the development path that is more suitable for themselves at this stage based on their actual situation and needs.

 

(