Please note: The algorithm descriptions in English have been automatically translated. Errors may have been introduced in this process. For the original descriptions, go to the Dutch version of the Algorithm Register.
Github Copilot
- Publication category
- Other algorithms
- Impact assessment
- DEDA
- Status
- In use
General information
Theme
Begin date
Contact information
Responsible use
Goal and impact
GitHub Copilot is an AI-enabled code assistant that helps developers by automatically making code suggestions during programming. The goal of Copilot is to increase productivity, reduce routine tasks and speed up the learning process. Copilot works as a plug-in in popular development environments and analyses the context of the current code, the project and the programming language used. Based on this, it generates relevant suggestions, ranging from small code snippets to complete functions or algorithms.
Copilot's suggestions are based on a large number of publicly available code examples and best practices. As a result, it can not only provide standard solutions, but also suggest alternative approaches and optimisations. Developers can directly adopt, modify or ignore the suggested code.
Copilot supports many different programming languages and frameworks, making it widely applicable in various software projects. It helps in writing boilerplate code, discovering new programming patterns and solving problems faster. Despite the smart suggestions, the developer remains responsible for checking, testing and securing the final code. Copilot is thus a powerful tool that makes the development process more efficient and informative, without taking control out of the developer's hands.
Considerations
GitHub Copilot uses artificial intelligence to generate code suggestions based on the context in your editor. It analyses your current code and input, sends it (anonymised and without private data) to the Copilot cloud service, where a trained AI model generates relevant suggestions. These suggestions are sent directly back to your editor. Data processing is done securely: Copilot does not use private code for training without explicit permission and processes user data according to strict privacy guidelines.
At the GitHub organisational level, we have not given permission to process our user data.
Human intervention
Code generated by GitHub Copilot is always checked by a programmer. Code published in the main branch is always verified.
Risk management
Generated code is always checked by means of a pull-request. A log is also kept of written and published code. Github Copilot is only used to optimise existing code or to generate new (parts of) code. No data, records or other data are passed to the model.
Impact assessment
Operations
Data
GitHub Copilot is trained on a large dataset consisting of publicly available source code from GitHub repositories and other publicly accessible text and code resources. Private code is used only if you have given explicit permission. The training data includes many programming languages and frameworks to provide relevant code suggestions.