.jpg)
Phase 1:
Explore the incubation period
Agriculture and energy are representative industries. There are still relatively few institutions in the industry that try to adopt large models, but there are still some leading or innovative institutions actively exploring.
The key for institutions to promote market applications is to prove the feasibility and practicality of the technology, and to be able to solve industry-specific challenges, face higher risks and uncertainties, and have the opportunity to lead the market. For example, in the power industry, state-owned power grid companies are actively trying, but due to the high requirements for key indicators such as safety, the promotion is cautious and the speed is relatively slow.
Phase 2:
Accelerated test period
Education, finance, games and travel are representative. The industry generally has a relatively good data foundation, and the number of institutions exploring the application of large models is growing rapidly, and they are beginning to generate economic value in specific application scenarios.
Institutions focus on how technology can solve practical problems, such as the winning rate of financial quantitative strategies, cost reduction and efficiency improvement of game design, etc. Success stories are the vane of this stage, and practical benefits can attract more participants to join.
Phase 3
Adoption growth period:
The advertising and software industries are representative. Mainstream organizations in the industry have generally adopted and used big models.
Due to its high compatibility with the basic capabilities of big models, capabilities such as copy generation, text graphics, code generation and data analysis have been widely used in many organizations in the advertising and software industries [including various Internet applications].
The key to continued market expansion lies in further optimizing technology applications, improving user experience and efficiency, while reducing costs
Phase 4
Landing maturity stage:
This stage means that the application of large model technology is basically mature, and most organizations have used it in major production and operation scenarios, and have established stable business cooperation relationships with suppliers.
At present, the big model technology itself is far from mature, and it will take longer for industry applications to mature. The stability, interpretability, and reliability of plug-in calls of big models are all necessary prerequisites for industry applications to enter the mature stage .
Industry AI Applications
Three main functions of AI
1) Content generation and creative design
It mainly uses the generation capabilities demonstrated by large models, including text generation, image generation, and pan-text generation capabilities such as codes and tables, combined with data from specific industries and scenarios to support content generation and creative design.
For example, all industries generally require marketing and advertising. Whether it is copywriting in the creative stage or advertising material generation in the market stage, the application of advertising big models combined with industry-specific data can play a better role.
In the wave of digitalization, many industries also need to increase code development, data analysis and other work. Big models assist in code generation, data analysis and chart generation, etc., which have become a common demand across industries in growth.
2) Information extraction and professional assistance
It mainly uses the summarization and planning capabilities of large models to assist people in extracting, analyzing and processing professional knowledge for data in specific industries and scenarios. Combined with technologies such as search enhancement generation, many industries have realized such assistant applications through dialogue robots, covering multiple links such as R&D design, production and manufacturing, and marketing services.
For example, in the financial industry, the financial big model formed by combining the proprietary knowledge in the financial field can effectively support the front-end and back-end work, and become the key support for financial investment decisions, risk assessment, etc.; in the medical industry, scientific researchers can use the interactive drug development assistant to conveniently query professional information, help with new drug development, and even assist in DNA sequencing to identify the relationship between diseases and genes.
3) Task scheduling and intelligent interaction
The industry's demand for big models also reflects expectations for their agent capabilities. It is hoped that big models can be connected with other applications and even with machines and equipment in the real world to assist in task scheduling and problem solving in a wider range.
This involves real-time data processing, automated control, environmental perception, and decision support, which places higher demands on the response speed, accuracy, and adaptability of the model, and requires the smooth development of large model plug-in ecosystems, the combination of large and small models, and multimodal large models. For example, the power industry hopes to effectively integrate sensor data through the task scheduling and interaction capabilities of large models, optimize energy distribution and consumption, and improve operational efficiency in the energy field; the transportation industry involves complex real-world interactions and has high requirements for safety and accuracy. It may be necessary to develop a native transportation large model with multimodal capabilities to meet the needs.
AI application success criteria
Based on the research of multiple practices and combined with relevant international frontier explorations, we try to summarize and construct the 2-3-1 principle for measuring the success of the application of big models in the industry:
Avoid two misunderstandings:
Evaluate three types of value,
Build a pattern.
Evaluate three types of value
1) Reduce costs and improve efficiency
Cost reduction and efficiency improvement mainly measure the impact of the industry big model on the operation and management of the organization itself, and it is also the most obvious performance at present. The core is that the big model can help enhance the ability of personnel, improve the level of automation and simplify the process, and ultimately reduce labor costs and improve the efficiency of organizational operation and management.
Typical scenarios include automated development, business process optimization, intelligent decision support, etc. Indicators for measuring effectiveness include cost savings, time savings, efficiency improvements, and the accuracy of decision-making support provided by large models.
→ 1.1 Automated development
The large model can automatically generate code and provide programming suggestions, reducing developers' repetitive work and allowing them to focus more on complex system design and problem solving. At the same time, this also lowers the programming threshold, allowing non-professionals to participate in programming development and promote the popularization and innovation of technology.
For example, the code examples and suggestions provided by the big model help developers avoid common errors and improve efficiency. The big model can understand the syntax and structure of the programming language and even the specific context of the project, providing customized support for developers.
This automation not only speeds up the development process, but also improves product quality.
→ 1.2 Business process optimization
Big models can be embedded in business process links through intelligent assistants and other means to improve the efficiency of interaction between links, and simplify and optimize processes through automation to improve business operation efficiency.
For example, in the gaming industry, large models can quickly transform initial concepts into visual content, reduce communication costs between planners and artists, and improve game development efficiency.
In the energy industry, such as power grid operation and maintenance, large models combined with vector databases accelerate the creation of knowledge graphs, support data processing and decision-making process efficiency, and improve operational standards and safety.
In the financial industry, big models have been used in automated trading, risk management, and compliance monitoring to assist in identifying risks such as fraud while improving the speed and accuracy of transaction processing.
Big models are also being incorporated into the collaborative office space. For example, video conferencing combined with big models can automatically generate meeting records and summaries, form action plans, and automatically distribute task information.
→ 1.3 Intelligent decision support
Big models can also play a role in operational analysis and management. For example, in supply chain management, big models can assist in analyzing and predicting demand, support inventory optimization, improve logistics efficiency and reduce costs.
In the field of transportation, big models can integrate multiple data such as people, vehicles, roads, and the environment to provide intelligent control solutions and improve the flow management of urban traffic. In the financial industry, big models can assist in efficient information retrieval and improve the decision-making and service efficiency of wealth management consultants.
In terms of data analysis, large models can assist in data exploration, automatically perform tasks such as data cleaning and pre-processing, save analysis time, and improve the timeliness of analysis results.
The advertising industry, based on the big model, uses real-time data feedback from delivery to drive the generation of creative materials and the setting of various indicators in budget allocation, and ultimately establishes a clear growth model and delivery strategy.
2) Business Innovation
Business innovation mainly measures the impact of the industry's big model on business supply capabilities, and is also the goal that many organizations most want to achieve.
The key is that the ability to generate large models can expand content supply, and its combination with application scenarios may also create new functions or businesses, helping to improve business competitiveness and expand business market space.
Typical scenarios include enriching content creativity, optimizing business functions, developing new businesses, etc. Indicators for measuring effectiveness include the speed and quality of content generation supported by the big model, the number of business users, and business revenue.
→ 2.1 Lowering the threshold for creativity
Large models can generate content quickly, thereby lowering the threshold for content creativity, especially in the fields of advertising, architecture, planning and design.
The advertising and marketing industry is the most typical example. Technological advances, especially the continuous iteration and upgrade of large-scale visual images and videos such as Midjourney, Stable Diffusion, Sora and Pika, have the most direct impact on advertising.
These models make creativity no longer an unattainable resource. Ordinary people can also use AI tools to visualize their ideas, greatly enriching the supply of advertising creativity. For example, in the architecture and planning industry, large models can run through the stages from creative initiation to review, assisting in gradually transforming creativity into reasonable designs.
→ 2.2 Enriching creative supply
First, it will expand the supply. As the threshold for creative generation is lowered, ordinary people can also present their creative ideas through AI tools and carry out various reprocessing and applications. The hybrid approach of AIGC+UGC will greatly increase the supply of creativity.
The second is to enrich the supply types. Currently, Wenshengtu is only the 1.0 version of AIGC. With the emergence of new technologies such as Sora, the imagination of "throwing a novel to generate a picture book or a movie" is becoming a reality.
In the future, Vincent text will be combined with Vincent pictures, Vincent videos, game engines, etc. to create more content forms and service methods.
3) Experience Enhancement
Experience enhancement mainly measures the impact of industry big models on user usage and reflects the value to users. Big models can provide natural language interaction capabilities, and as they develop towards multimodal and embodied intelligence, they provide users with a more natural and rich experience, thereby creating value-added.
Typical scenarios include changes in interaction methods, personalized services, virtual companionship, etc. Indicators for measuring effectiveness include user usage, activity, satisfaction, problem resolution rate, retention rate, etc. of the big model-supported business.
3.1 Changes in interaction methods
The big model may change the way existing applications interact, allowing users to use applications and call functions in a more natural conversational way.
For example, documents can be combined with large models, and users can describe their needs and let the application directly generate text, graphics and other content. What is more imaginative is the combination with new interactive technologies and devices such as AR/VR, such as Apple's Vision Pro, which may promote a new round of changes in application forms and experiences.
3.2 Personalized Service
Big models can provide personalized services. For example, in the field of education, big models can provide personalized teaching content and services for each student based on their characteristics and interests.
At the same time, in the teaching process, dialog learning will also enhance the interactivity of students' learning process and improve each student's interest in learning. In the medical field, large models can provide assistant services for doctors and patients, and assist in the implementation of large-scale precision medical services.
3.3 Virtual Companion Service
In the aspect of product service optimization, integrating the capabilities of big models into products has become the focus of exploration for improving the intelligent capabilities of products in consumer electronics, automobiles and other fields.
For example, the smart speaker Vifa ChatMini has built-in ChatGPT, which improves the natural language interaction experience while maintaining professional acoustic standards. It can be used for emotional support, learning companionship, work assistant, etc.
Comentários