.png)
HOCIntelligentTechnologyGroup
Faculty, BIG DATA (information), Large language model(LLM), Generative Pre-trained Transforme model(GPT), Research Lab's "FBLGR Quality" is a leading company,AI Science and Technology Innovation for Global Sustainable Development.
DX, ICT, AI, Agriculture, Medical, Human Resource Development, Future Prediction, SDGs, etc.
The virtual software development company Services
The virtual software development company generates an AI GPT company，code and documentation one-stop service
We all know that large language models (LLMs) represented by ChatGPT have code generation capabilities, after all, code itself is also a language. All employees of the company are LLM, which can complete the entire software development process from analyzing requirements to writing code to document production end-to-end, and realize one-stop software development services. Ideally, based on this framework, users can simply ask for a request and get a well-documented software! Involves software requirements, design, implementation, testing, and maintenance. The software development process involves a variety of job functions, including organizational coordination, task assignment, code writing, system testing, documentation, and more. Therefore, complex software requires long development cycles and a high level of attention to detail.
a model framework running in a company mode, and divides the development process into four stages: design, writing code, testing and documentation. Several different agents are used at each stage, which play different roles in the company, such as programmers, reviewers, and testers. To facilitate effective communication and collaboration, each phase is divided into atomic, easy-to-perform subtasks. In a chat chain, each node represents a specific subtask, and two roles engage in multiple rounds of context-aware discussions to propose and validate solutions. This approach ensures that customer requirements are analyzed, creative ideas are generated, prototype systems are designed and implemented, potential problems are identified and resolved, debugging information is interpreted, attractive graphical interfaces are created, and user manuals are generated. Guiding the software development process through a design chat chain along the chat chain (blockchain and chat) enables the delivery of the final software to the user, including source code, dependency specifications, and user manuals. is a chat-based software development framework. Users simply specify a task, and ChatDevelopment designs, writes code, tests, and documents sequentially. This new paradigm unifies the main development process through language communication, eliminating the need for specialized models at each stage, which greatly simplifies the software development process.
Started a service to generate legacy system design documents from source
Automated programming using large-scale language models
System development method using generative AI is rolled out to all engineers, man-month contracts are reviewed
Full-scale introduction of generative AI (artificial intelligence) in system construction for corporations. System development methods and tools using generated AI will be developed in-house and rolled out to development departments in Japan and overseas from FY2024.
There are about 10,000 engineers involved in business system development in Japan and overseas. We aim to have all engineers learn development methods that use generative AI and create a system that can be used by all employees, including related occupations such as sales.
In the demonstration, there have already been major achievements, such as a 70% reduction in development man-hours. If the productivity of system construction is greatly improved, it is possible to increase the number of projects that can be ordered, but there is also a contradiction that the unit price of projects and the unit price of customers may decrease. Because the order amount for system construction is mainly a "man-month type" contract that accumulates the necessary man-hours and uses it as a calculation basis, "we will organize the issues and consider new contract forms with customers, such as performance-based fee-based contracts," in preparation for the impact on the system construction business brought about by generative AI. Agile development reduces man-hours by 70%
On March 4, 2023, we launched Global Generative AI LAB, a global organization that promotes the use of generative AI. The organization will promote the development of generative AI services for customers and develop system development methods that use generative AI.
In the new system development method, generative AI is introduced in multiple processes. Highly effective has already been confirmed in the processes of "programming" to generate code and "unit testing" to verify the operation of the developed program on a functional basis. For example, in programming, a demonstration project applied to agile development had the effect of reducing programming work by 70%. The reason why it works well with Agile is that as you repeatedly code while adding and modifying functions, the range of code that can be automatically generated by AI and the accuracy of the generated code increase.
Positioning of existing system development methods and development methods corresponding to new generative AI
In addition, we will verify how generated AI can be used in other processes such as requirement definition and comprehensive testing. For example, in the upstream process, when documenting requirements definitions from customer requests and outline designs from requirement definitions, we believe that there is a possibility that it is possible to check with high accuracy whether there are any omissions in requirements and functions by using generated AI.
In addition, in the test process, a migration project that renovates business systems with new technology is listed as one of the candidates for easy application of AI. The "current new comparison test" to check whether the current system and the new system operate the same has the potential to be automated by AI with high accuracy.
Utilizing generated AI for data management operations, high affinity with data processing
We will cover data processing using generative AI (artificial intelligence) such as ChatGPT. Data processing has a high affinity with generative AI, and when verified by the group, it demonstrated high processing accuracy in various situations such as pointing out data quality problems, cleansing, name matching, anonymization, and format conversion.
One of the unique strengths of generative AI is that it can be used to output the "certainty" of the proposed amendment at 0~100%. You can set a threshold and automatically adopt amendments if they are above a certain level.
There are two aspects to the impact of generative AI on data processing. One is to broaden the scope of correction and improve data quality. In existing tools, there are products that incorporate a function that explicitly presents fluctuations in types, tags, notation, etc., but even if it is not implemented as a tool function in generative AI, it may be possible to respond if you instruct it. There is a possibility that the work of "making decisions" that people are responsible for can be greatly reduced. The other is the involvement of non-engineers in data processing. Business staff can be involved in data processing
When working with data with interactive generative AI, you only need to give instructions in natural language, and you don't need to learn how to use analysis tools. Flexibility is also a strength. You can freely divide categories such as "Support" and "Delivery", change the framework of personality analysis, and extract other data items.
