GitHub announced that it has completed testing of its intelligent assistant, GitHub Copilot, which can generate generic builds while writing code. The system was developed in collaboration with the OpenAI project and uses the OpenAI Codex machine learning platform, trained on a wide variety of source code hosted in public GitHub repositories.
Code generation supports Python, JavaScript, TypeScript, Ruby, Go, C#, and C++ using various frameworks. Modules are available to integrate GitHub Copilot with Neovim, JetBrains IDE, Visual Studio, and Visual Studio Code.
Judging by the telemetry collected during the testing process, the service allows for the generation of sufficiently high-quality code ; for example, developers accepted 26% of the recommendations proposed in GitHub Copilot as is.
GitHub Copilot differs from traditional code completion systems in its ability to form fairly complex blocks of code, up to synthesized ready-to-use functions taking into account the current context.
GitHub Copilot adapts to the way a developer writes code and takes into account the APIs and frameworks used in the program. For example, if there's a sample JSON structure in the comment, when you start writing a function to parse this structure, GitHub Copilot will offer ready-to-use code, and by writing routine enumerations of repeated descriptions, it will form the rest.
In a blog post, GitHub CEO Thomas Dohmke said that GitHub Copilot was designed as an extension of the editor to ensure that nothing gets in the way of what developers are doing.
“GitHub Copilot distills the collective knowledge of developers around the world into an editor extension that suggests code in real time, to help you stay focused on what matters most: building great software,” he explained.
According to Dohmke, around 1,2 million developers have tried Copilot during its preview stage. Apparently, it has also been quite useful, as Dohmke claims that he has written up to 40% of developer code written in popular languages like Python.
"Like the rise of compilers and open source, we believe AI-assisted coding will fundamentally change the nature of software development, giving developers a new tool to write code more easily and quickly," said Dohmke. .
GitHub Copilot's ability to generate pre-built blocks of code has generated controversy over possible infringement of the copyleft license. When forming the machine learning model, real source texts from open project repositories hosted on GitHub were used.
Many of these projects are provided under copyleft licenses, such as the GPL, which requires that the code in derivative works be provided under a compatible license. If Copilot pastes existing code, developers may inadvertently violate the license of the project from which the code was borrowed.
It is still unclear whether work generated by a machine learning system can be considered a derivative . Questions also arise about whether a machine learning model is subject to copyright and, if so, who owns these rights and how they relate to the rights of the code on which the model is built.
On the one hand, the generated blocks can repeat text passages from existing projects, but on the other hand, the system recreates the code structure and does not copy the code itself.
According to a GitHub study, only 1% of the recommendations suggested by Copilot include code snippets from existing projects longer than 150 characters. In most situations, this repetition occurs when Copilot fails to correctly determine the context or provides generic solutions to a problem.
To avoid replacing existing code, a special filter has been added to Copilot that does not allow overlapping with existing projects. When configuring, the developer can turn this filter on or off at his discretion. Among other problems, it is noted the possibility that the synthesized code could repeat the errors and vulnerabilities present in the code used to train the model.
Finally, it is worth mentioning that the service is free for maintainers of popular open source projects and for students. While for other categories of users, access to GitHub Copilot is paid ($10 per month or $100 per year), but free trial access is provided for 60 days.
Finally, if you are interested in learning more, you can find the details at the following link.