github announced on the completion of intellectual testing Assistant github copilot , capable of generating typical designs when writing code. The system was developed in conjunction with the Openai project and uses the Openai Codex machine learning platform, trained on a large array of the source texts posted in the GITHUB public repositories. The service is free for accompanying popular open projects and students. For other categories of users, access to GitHub Copilot is paid ($ 10 per month or $ 100 per year), but free introductory access to the course of 60 days is provided.
The generation of code in the programming languages Python, JavaScript, Typescript, Ruby, GO, C# and C ++ using various frameworks. The GITHUB Copilot integration modules are available with the development environment of Neovim, Jetbrains Ides, Visual Studio and Visual Studio Code.
Judging by the service collected in the process of testing telemetry, the service allows you to generate a sufficiently high quality code – for example, 26% of the recommendations proposed in GitHub Copilot were adopted by developers as it is.
The GitHub Copilot code is possible from the traditional auto -filling systems for the formation of quite complex code blocks, up to ready -made functions synthesized from the current context. Github Copilot adapts to the code of writing code by the developer and takes into account the API and frameworks used in the program. For example, if there is an example of a JSON structure in the commentary, when writing a function for analyzing this structure, GitHub Copilot will offer a ready -made code, and when writing routine transfers of repeated descriptions will form the remaining positions.
The ability of Github Copilot to generate ready -made code blocks led to disputes associated with a potential violation of the license copleg. In the formation of a machine learning model, real source texts from the repositories of open projects posted on GitHub were used.
Many of these projects are supplied under coplets with licenses, such as GPL, requiring the supply of work work under a compatible license. In the case of inserting the existing code proposed by Copilot, developers can involuntarily violate the license for the project from which this code was borrowed.
Can a derivative of the work generated by the machine learning system, so far is not clear