Possibility of AI used for coding-generation, understanding, application modernize-ZDNET JAPAN

　Can artificial intelligence (AI) understand the context of computer programs and generate code on their own?The advancement of the long legs to that goal promises a future that will make the work of developers and the work of non -developers with low code / no -code more productive and concentrate on the work at hand.is doing.

　In 2020, Intel learns the structure of the software code in collaboration between the Massachusetts Institute of Technology (MIT) and the Georgia Institute of Technology, and analyzes the differences in syntax from another code that performs similar movements.Announced the development of automated engines to learn the intentions of software."In the democratization of software production," said Justin Gottschlich, a principal scientist of Intel, and said, "When the machine is completely realized, everybody is intended.By simply expressing it, it will be possible to create software by the person's preferred way of proceeding, regardless of whether it is a code, a natural language, or something else. "

　Openai's natural language generation model "Generative Pre-Trained Transformer 3" (GPT-3) can also be used to automatically generate computer code. According to Vincent Tabora in an article by AI-related website, Vincent Tabora, the development manager will "establish a project by building an application base using a tool equipped with a GPT-3 function." He wrote, "You can use it as a skeleton to build the remaining part of the application. The manager only enters the request, produces the tools that require the manager. The project is the initial of the application. You will not have to arrange additional staff for construction. And after the skeleton is generated, the developers with advanced skills are introduced to implement other requests in the project, and applications are used. The procedure is to complete. "

　AI coding can not only be a means for building new applications, but also a means to smoothly shift applications and systems to modern platforms. In a recent interview, Maja Vukovic, an IBM Fellow of IBM's Thomas J. Watson Research Institute, struggles to shift over 1 million lines to new platforms by using AI. He introduced the case of supporting the client he was. He said, "The client's mission -critical applications have been enlarged by over 1.5 million lines." It was. In addition, the evolution of these code was promoted by multiple development teams, and some team members were moving to another department or even retired. In some cases, the document was not left. "