Recently, the People’s Daily Finance Research Institute released the report “Opening a New Era of Intelligence: 2024 China AI Model Industry Development Report”, analyzing and comparing the functionality and applicability of eight AI models with high market share in China.
iFlytek Spark Cognitive ModelThe iFlytek Spark Cognitive Model has seven core capabilities:text generation, language understanding, knowledge Q&A, logical reasoning, mathematical abilities, coding capabilities, and multimodal capabilities.In knowledge learning and content creation, it can performelement extraction, question generation, helping to create richer and more useful intelligent agents in the fields of knowledge learning and content creation, and reasonably expand by integrating external knowledge.
Applicability AnalysisUsers can usetext generationto experience one-click document generation, AI writing assistants, multilingual document generation, AI automatic image matching, multiple template selections, speech notes, and more. By asking questions, users can obtain knowledge about daily life, medical knowledge, policy interpretations, and more.The iFlytek Spark cananalyze the premises and assumptions of problems to infer answers or solutions, providing new ideas and insights. In research tasks, it can use existing data and information for inference, prediction, and verification. It can also solve mathematical problems such as equation solving, solid geometry, calculus, and probability statistics.In terms ofcoding capabilities, it can intelligently generate code based on comments and function names, supports line-by-line code comments, can accurately locate syntax and logical errors in code, and can even intelligently generate unit test data.In terms of multimodal capabilities, the iFlytek Spark can return accurate image descriptions based on user-uploaded images or answer questions related to image materials. It can also generate desired audio and video based on user descriptions.
Wenxin Yiyan ModelWenxin Yiyan can understand subtext, complex sentence structures, professional terminology, disordered sequences, and vague intentions, and is also capable of code understanding and debugging tasks.Applicability AnalysisWenxin Yiyan can be applied to the creation of literary works such as novels, essays, and poetry.In terms of copywriting, it can write business plans, market analysis reports, and other commercial copy. It provides advertising creative thinking, inspiration, and ideas, quickly coming up with attractive advertising copy and slogans; the Wenxin Yiyan chatbot is applied in life services, educational tutoring, customer service, and other fields.In terms of multimodal generation, the Wenxin model supports image generation and processing, can generate images based on user needs or edit existing images. The Wenxin model also supports speech synthesis, speech recognition, and audio classification. The Wenxin model can also process video data or convert text into dynamic image sequences to complete video classification, object detection, and other tasks.In terms of mathematical logical reasoning, the Wenxin model can solve complex mathematical problems and can also serve as a coding assistant. For example, Baidu has developed an intelligent coding assistant called Comate based on the Wenxin model, providing smart recommendations, intelligent generation, and intelligent Q&A features, supporting multiple programming languages and IDEs.In terms of generation capabilities, it can quickly generate texts, codes, images, charts, and videos in various styles, such as copywriting, creating life plans, and writing high-quality code.In terms of logical capabilities, it can help users solve complex logical problems, difficult mathematical calculations, important career/life decisions, code corrections, common sense reasoning, logical verification, solid geometry, debate inspiration, and more.
Memory AbilityIn terms of memory ability, after multiple rounds of dialogue, Wenxin Yiyan can still remember the key points of the conversation, easily handling complex problems and immersive role dialogue.
Tongyi Qianwen ModelTongyi Qianwen can provide users with rich interactive experiences in creative copywriting, office assistance, learning assistance, and fun life aspects.Applicability AnalysisCreative Copywritingapplications include: “writing marketing copy”, entering product descriptions to obtain customized gold medal marketing copy. “Article Polishing” can deeply analyze the articles submitted by users, exploring the inadequacies in expression and providing suggestions for vocabulary and sentence variations. “Live Streaming Sales Script Generation”, based on rich product information and user needs, provides e-commerce anchors with vivid and interesting script content with marketing power.Office Assistantapplications include: “SWOT Analysis” provides users with comprehensive, in-depth, and precise strategic decision-making support, understanding and assessing the impact of internal and external environments on specific projects from multiple perspectives. “PPT Framework Generation” intelligently constructs a professional and logically clear PPT structure for users.Learning Assistantapplications include: “Question Processing Plant”, which generates high-quality test questions based on the provided specialties and subject areas, greatly saving the time and energy of teachers, parents, and educational institutions in question creation. “Learning Plan Station” provides users with personalized, systematic learning path planning, customizing efficient and scientific learning schedules.Fun Lifeapplications include: “Recipes that Fly Away”, guiding users step by step to provide cooking secrets for delicious dishes. “AI Fitness Coach” creates personalized fitness plans for users. “Lyric Writing” generates vivid lyrics based on the song titles provided by users.
Chitu ModelThe Chitu Model is a vertical industry multi-level large language model developed by Ronglian Cloud for enterprise applications, empowering enterprises to build dedicated intelligent customer service and intelligent marketing, including conversational insights, business scripts, Q&A knowledge bases, knowledge application, data analysis, intelligent dialogue frameworks, and process management. The three core points areintelligence,controllability, andcost-effectiveness.Applicability AnalysisBased on the Chitu model, Ronglian Cloud released the generative application “Rongxi Copilot”.Large Model Scripts:Rongxi Copilot can quickly verify and filter massive historical conversation data with one click in the background, selecting better scripts and generating gold medal scripts, balancing quality and quantity while uncovering frequently concerned issues from customers, gaining insights into business pain points.Intelligent Knowledge Base:It can help enterprises quickly build a script library from scratch at low cost, including understanding document knowledge, knowledge quick search, intelligent Q&A, etc., significantly improving construction efficiency.Conversational Insights:Efficiently and conveniently gain insights into every conversation communication situation, analyzing customer demands, precisely diagnosing problems, and optimizing. Returning to the actual business itself, Rongxi Copilot deeply integrates into the financial industry’s segmented scenarios, creating scenario-based customer service assistants, such as installment retention assistants, card recommendation retention assistants, complaint appeasement assistants, etc., providing real-time assistance in quickly understanding customer needs, recommending better response scripts, diagnosing customer emotional changes, and reminding wording and precautions.
Wenxiu ModelThe Wenxiu model provides proofreading services more suited to the usage scenarios for professional users such as government agencies, news media, enterprises, educational institutions, and publishing houses. It hasstrong proofreading capabilities, fast speed, and high matching degree as three major characteristics, better solving problems in vertical industries.Applicability AnalysisGovernment UnitsIn the field of government units, it empowers various levels of government departments to automate proofreading processes, providing content error sensitivity proofreading, modification prompts, and text polishing services, ensuring the accuracy and rigor of content while supporting proofreading in intranet environments to meet higher confidentiality requirements.News Media FieldThe Wenxiu model deeply integrates into various aspects of news media work, performing multi-type error sensitivity proofreading on multimodal content, helping to quickly locate errors and highlight them, making content more standardized and rigorous, effectively maintaining the credibility of official accounts; while also providing text polishing services to improve publication speed, ensuring news timeliness.Enterprise Units FieldIt fully engages in enterprise office scenarios, optimizing promotional content from multiple aspects such as content error correction and improving text quality, enhancing the attractiveness of copy and significantly improving marketing effectiveness.Educational Institutions FieldFor educational institutions, it conducts comprehensive reviews of promotional materials, new media articles, research reports, academic papers, etc., effectively reducing text error rates and ensuring academic rigor. Through the AI polishing function, it assists in drafting and optimizing articles and reports, helping to further enhance the school’s communication and influence; in the publishing institution field, it provides professional, convenient, and efficient content screening and text quality assurance services, assisting various publishing institutions in efficiently processing multilingual texts, reducing content error probabilities, and ensuring content standardization and accuracy.
YonGPT ModelThe YonGPT model’s applications in the enterprise service field mainly focus on four directions:business operations,human-computer interaction,knowledge generation, andapplication generation.Applicability AnalysisIn terms ofintelligent business operations, YonGPT uses data analysis and prediction capabilities to gain deep insights into enterprise operations, identify potential business risks and opportunities, and provide intelligent solutions to improve decision-making levels and operational efficiency.In terms ofnatural human-computer interaction: YonGPT enables enterprise applications and services to engage in natural and smooth dialogues with users through powerful natural language processing technologies and understanding capabilities, achieving the invocation, connection, and assembly of different applications in a human-centered way, completing work more naturally and efficiently.In terms ofintelligent knowledge generation: YonGPT extracts and integrates knowledge from massive data and information, generating new and valuable knowledge content, covering industry solutions and professional knowledge sharing, helping enterprises and users fully utilize their knowledge reserves and accumulation, promoting knowledge dissemination and application.In terms ofsemantic application generation: YonGPT can automatically generate applications with semantic capabilities by understanding user needs, enterprise business, and data characteristics, greatly enhancing the efficiency of creating personalized application services for enterprises.YonGPT’s intelligent scene services include four services:Intelligent Analysis of Business Revenue/Profit and Tax Operations, which can monitor operational conditions in real-time, quickly identify issues, accurately predict enterprise benefits, and effectively foresee changes.Intelligent Order Generation, which integrates rich supply chain experience, realizes rapid intelligent order generation through an “interactive innovative” order generation assistant, improving enterprise efficiency.Intelligent RecruitmentThrough AI interaction, it optimizes the job application experience, achieving precise decision-making in selecting and using personnel.Intelligent Big Searchprovides an “immersive” new search experience, accelerating the value realization of enterprise knowledge services, gaining insights into user needs, and achieving integrated search and push, empowering business and organizations with knowledge.
“Xieyi” Intelligent Creation EngineThe “Xieyi” Intelligent Creation Engine is suitable for groups with daily reading and writing needs such as party and government media, central enterprises, state-owned enterprises, schools, and hospitals. The “Xieyi” Intelligent Creation Engine deeply explores user needs, building an interactive experience of “searching, writing, and reviewing”, characterized by high efficiency in creation, safety, accuracy, and rich content.Applicability AnalysisThe People’s Daily customizes the training of the “Xieyi” intelligent creation language model for customers based on the industry client database corpus, supplemented by the content of the People’s Daily, helping to improve written expression capabilities, accumulate writing materials, and standardize writing formats. During the writing process, it can also provide rich materials for title writing, rhetorical usage, quotations from poetry and literature, and internet slang, helping creators inspire ideas and provide thoughts, thus automatically and efficiently generating high-quality article materials that meet the writing scenarios of customers, providing comprehensive, timely, and all-dimensional intelligent services, greatly enhancing overall work efficiency.
Lanxin ModelThe Lanxin model is the first open-source self-developed large model running on mobile terminals in the industry and is more suitable for Chinese users. With the increase in parameters, the Lanxin model gradually possesses capabilities such as text summarization, language understanding, text creation, knowledge Q&A, role-playing, complex logical reasoning, and complex task arrangement. Based on the capabilities of the Lanxin model, vivo has developed two mobile terminal products: Lanxin Xiao V and Lanxin Qianxun.Applicability AnalysisLanxin Xiao Vsupports semantic search, Q&A, writing, image creation, and intelligent interaction. The super image creation function includes: text-to-image and image-to-image, AI bystander elimination (bystander invisibility).Lanxin Qianxuncovers two core application scenarios: AI dialogue and AI inspiration. Lanxin Qianxun is the first publicly available free APP of a large model in the mobile industry. Lanxin Qianxun can provide functions such as social media copywriting, PPT outline generation, and Chinese-English text translation, and also has tools such as dressing suggestions.
Source: China Science and Technology Information