At the beginning of this year, the domestic artificial intelligence (AI) large model DeepSeek became popular for its innovation. Representing cutting-edge technology, AI large models like DeepSeek are becoming a key element for developing new productive forces. Capitalizing on the opportunities in AI development, units in the tobacco industry are actively pursuing the localization deployment or upgrade of DeepSeek large models, exploring applications in various scenarios and fields, and developing abilities in AI perception, AI decision-making, and human-machine interaction to deepen the empowering effect of AI technology on business development.
Today, let's take a look at the exploration and implementation of the localization deployment of the DeepSeek large model in some units of the tobacco industry.
The Guangxi Zhuang Autonomous Region Tobacco Monopoly Bureau (Agency) recently upgraded the AI large language model DeepSeek from version V2.5 to version R1, planned for deployment in October 2024, to provide more efficient and accurate intelligent services for all system staff.
The latest upgrade of DeepSeek further enriches and expands five major scenarios of the "Guiyan AI Assistant": intelligent Q&A, intelligent data questioning, compliance review, writing assistant, and analytical reports. The "intelligent Q&A" integrates internal system data such as official document systems and comprehensive management platforms, while also incorporating external information sources to provide higher quality answers in policy interpretation, business processes, and operational norms. The "intelligent data questioning" seamlessly connects 58 key data indicators and dimensions of marketing business, utilizing strong semantic understanding capabilities to accurately grasp users' data needs and display precise and detailed data charts. The "compliance review" expands the identification of potential risk points in various types of report materials and offers more rational improvement suggestions. The "writing assistant" gathers news resources from internal and external corporate websites, providing fresh wording, clearer logic, and more detailed details in writing tasks. The "analytical reports" conduct in-depth exploration and multidimensional analysis of enterprise data to expand 24 categories of marketing data analysis reports, providing a comprehensive and scientific basis for management decisions.
Recently, the Anhui Tobacco Business Law Popularization and Publicity Education Base has launched the first "smart and interactive" DeepSeek scene application, called "Little Tao Lawyer 2.0".
This application, targeting the robot "Xiao Tao Lawyer" in the legal education base, has been upgraded to version 2.0 through the transformation of adapting the DeepSeek large language model, plugging in the RAG knowledge base, customizing workflows, and integrating Internet data resources. After the upgrade, in addition to the existing legal consultation function, it has enhanced features such as voice recognition, knowledge retrieval, and situational understanding. This allows it to provide more refined and personalized communication, legal consultation, venue guidance, and other services for visiting guests, further enriching the form of exhibition explanations in the exhibition hall.
Exclusive sales of AI intelligent assistants
Recently, the "Specialized AI Intelligent Assistant" independently developed by the Tobacco Monopoly Bureau of Xi'an City, Shaanxi Province, has officially launched and started running. This intelligent law enforcement tool is equipped with the DeepSeek large model and relies on local knowledge bases and the national unified monopoly supervision platform to provide comprehensive intelligent support for monopoly enforcement officers.
After the AI smart assistant launched and began operating, law enforcement officers only need to ask questions through voice or text to the AI assistant, which can then access relevant data from the local tobacco monopoly supervision knowledge base to provide indexes and answers. This not only helps law enforcement officers improve their knowledge of the industry, but also greatly increases work efficiency. According to statistics, in traditional modes, it takes over 10 minutes on average to search for legal provisions and document references, but with the AI smart assistant, this time is shortened to less than 10 seconds, increasing efficiency by over 60 times. Additionally, depending on the circumstances, the AI smart assistant can provide personalized enforcement process guidance for law enforcement officers, effectively reducing procedural errors and operational mistakes, ensuring that law enforcement actions are standard and legal.
Automatic cigarette pack scanning system.
Recently, the "Cigarette Pack Automatic Scanning System" jointly developed by Hunan Tobacco Re-roasting Co., Ltd., Changde Tobacco Monopoly Bureau (Company), and Xiangxi Tujia and Miao Autonomous Prefecture Tobacco Monopoly Bureau (Company) has successfully passed internal testing. This project innovatively integrates the DeepSeek big model with the tobacco scanning business, successfully overcoming the technical challenges of dynamic scanning technology.
The project team fully leveraged the DeepSeek large model deep learning and efficient data processing capabilities, combined with cutting-edge technologies such as artificial intelligence, edge computing, and computer vision, to successfully build the industry's first library of cigarette pack and label shapes. At the technical level, they achieved precise separation, identification, and data verification of cigarette packs, labels, and QR codes in a mobile operating environment, greatly improving the efficiency and reliability of data processing. At the business level, they overcame the inefficiencies of traditional code patch operations, achieving precise identification and repair of missing or incorrect codes. By using AI recognition algorithms and computer vision technology, they intelligently analyze and detect violations such as package tampering, code switching, and label duplication during transport, significantly enhancing the accuracy and operational efficiency of the tobacco leaf sweeping code business.
The Little Prince of AI
Recently, the intelligent agent application "AI Little Prince" deployed by Hunan Zhongyan Industry Co., Ltd.'s DeepSeek localization department has been officially launched.
At present, the "AI Prince" can achieve intelligent question and answer in company management standards, management systems, department standards, etc., and set up data security protection measures. In the next stage, Hunan Tobacco will further strengthen the general service capabilities of large models, open the AI underlying platform to the entire company, and continuously explore integration points of AI technology with business scenarios, covering the company's main business lines, and forming an intelligent application scenery 2.0 map. By quickly building more intelligent body applications, promote the digital transformation of "tobacco + AI", support the company's "dual-new integration" strategy, empower high-quality development and modernization construction of enterprises.
Rolling bag repair knowledge assistant elf
On March 19th, the "Rolling and Packaging Maintenance Knowledge Assistant Elf" developed independently by China Tobacco Hubei Industrial Co., Ltd's Wuhan Cigarette Factory Rolling and Packaging Workshop was officially launched and put into operation. This tool integrates deeply with industrial systems through large language models, providing intelligent decision support for equipment maintenance and opening a new stage of AI-assisted maintenance.
According to reports, the platform was developed by the digital creator team in the rolling packaging workshop based on the DeepSeek-32B large model. By analyzing more than 13,000 equipment maintenance records from 2021 to 2024, a knowledge base covering five major categories of equipment such as winding machines and packaging machines was constructed. It innovatively adopts a dual-drive mode of "outboard knowledge base + large language model," achieving real-time response of fault diagnosis solutions. After the system went online, maintenance personnel can quickly query fault codes and access historical case handling through the workshop production information management system, and obtain AI-generated maintenance strategy optimization recommendations.
An Yan AI Assistant
Since the beginning of this year, the Yantobacco Monopoly Bureau (Company) in Anyang City, Henan Province has deployed the DeepSeek R1 distilled version model and the Dify generative AI application innovation engine based on actual needs. They have designed the "Anyan AI Assistant" intelligent body, using Embedding technology to transform text, images, audio and other data into low-dimensional vectors, constructing a local legal knowledge graph, calling on the DeepSeek model to parse key words in the problem and match them with the knowledge base, then reordering with Rerank technology, and finally having DeepSeek organize and optimize language output high-quality results, providing efficient solutions for content generation.
In practical applications, traditional compliance work relies on manual inspection of a large number of legal documents, which is not only cumbersome but also error-prone. Nowadays, the "Anyan AI Assistant" relies on powerful large-scale model reasoning capabilities and local legal knowledge graphs to stand guard in key areas such as procurement and engineering. Once any signs of violation are detected, warnings are immediately triggered. The intelligent review decision support system built on AI provides precise and reliable suggestions for standardizing and purchasing decisions, helping enterprises accelerate their digital transformation.
Lin Xiaoan" AI assistant
Ensuring safety in production is the lifeline of development. The tobacco monopoly bureau (branch) of Lihai City, Zhejiang Province has recently been following the trend of digital transformation. By leveraging the user-friendly low-code development platform and combining it with the DeepSeek large model, they have created the "Linxiaoan" AI assistant.
Linxiaoan" can achieve online full-process control of safety work. During daily inspections, grassroots safety officers can ask questions online to "Linxiaoan" if they encounter unclear inspection regulations or are unable to identify risks. "Linxiaoan" can also accurately push inspection forms to grassroots safety officers, who can fill them out online, check in with their location, upload photos, and report on the inspection situation, thereby improving the efficiency of daily inspections. At the same time, organizational safety administrators can efficiently follow up on the progress of daily safety inspections through the intelligent safety inspection reports generated by "Linxiaoan" instantly.
The first AI course in the new tobacco industry is set to begin, with high demand for enrollment.
On April 15th, the "First AI Course for New Tobacco Industry Workers", co-created by 2Firsts and Smart Dog AI, will soon begin in Qianhai, Shenzhen. As the industry's first AI course, it will help professionals embark on a journey of intelligent transformation. Registration is currently underway, seats are limited, first come, first served, so seize the opportunity!
Notice
1. This article is provided exclusively for professional research purposes related to industry, technology and policy. Any reference to brands or products is made solely for the purpose of objective description and does not constitute an endorsement, recommendation, or promotion of any brand or product.
2. The use of nicotine products, including but not limited to cigarettes, e-cigarettes, and heated tobacco products, is associated with significant health risks. Users are required to comply with all relevant laws and regulations in their respective jurisdictions.
3. This article is strictly restricted from being accessed or viewed by individuals under the legal age.
Copyright
This article is either an original work by 2Firsts or a reproduction from third-party sources with the original source clearly indicated. The copyright and usage rights of this article belong to 2Firsts or the original source. Unauthorized reproduction, distribution, or any other unauthorized use of this article by any entity or individual is strictly prohibited. Violators will be held legally responsible. For copyright-related matters, please contact: info@2firsts.com
AI Assistance Disclaimer
This article may have utilized AI to enhance translation and editing efficiency. However, due to technical limitations, errors may occur. Readers are advised to refer to the sources provided for more accurate information.
This article should not be used as a basis for any investment decisions or advice, and 2Firsts assumes no direct or indirect liability for any errors in the content.
