According to Maria Diaz of ZDNET, the utilization of generative AI in coding is a topic of debate. One experimental trial found OpenAI’s ChatGPT tool capable of producing quality code. However, other studies indicate that large-scale language models like GPT-4 fall below human coders in overall code quality.
Yet, some argue that the debate over AI’s abilities as a coder is not the main issue. They believe that coding assistance through automation is altering the nature of a programmer’s job. In a recent interview with ZDNET, Inbal Shari, Chief Product Officer for GitHub, emphasized the advent of an abstraction layer in the form of natural language. This layer has initially been used for code completion, although Shari insists its potential to extend beyond mere code completion.
In 2021, GitHub launched its code assistance tool known as GitHub Copilot, and it has garnered significant traction. For example, Accenture, a prolific user of Copilot, has reported a high retention rate of Copilot-generated code. This, in turn, has contributed to increased productivity in the completion of pull requests and the build process.
Shari suggests that the switching between different tools has been reduced by Copilot’s integration into a programmer’s IDE. This has resulted in greater developer satisfaction through decreased context-switching among tools. As GitHub plans to introduce an enterprise version of Copilot in February, the tool aims to adapt to individual developer’s coding style and cater to personalized needs.
The ongoing measurement of developer productivity, user satisfaction, and impact are vital aspects that need to be further understood in the context of AI’s application in coding. As GitHub continues its program development and looks to collaborate with customers, the journey to determining AI’s impact on coding productivity remains a work in progress.