GitHub Copilot is an AI tool that helps devs to write better code


Zach Marzouk

30 Jun, 2021

GitHub has launched a preview of GitHub Copilot, a new AI pair programmer that helps developers to write better code.

The Microsoft-owned company said the tool “draws context from the code you’re working on,” and suggests whole lines or entire functions. This, GitHub says, could enable engineers to be more productive and potentially lower the barriers to enter software development.

“It helps you quickly discover alternative ways to solve problems, write tests, and explore new APIs without having to tediously tailor a search for answers on the internet. As you type, it adapts to the way you write code—to help you complete your work faster,” said Nat Friedman, CEO of GitHub.

GitHub Copilot was developed in collaboration with OpenAI, an AI research startup backed by Microsoft, and powered by OpenAI Codex, a new AI system created by the company. Essentially, the GitHub Copilot editor extension sends comments and code to the GitHub Copilot service, which in turn uses OpenAI Codex to synthesise and suggest individual lines and whole functions.

The company said the system “has broad knowledge of how people use code” and is more capable than GPT -3 in code generation, as it was trained on a data set that includes a larger concentration of public source code.

GitHub expects the technology to “enable existing engineers to be more productive, reducing manual tasks and helping them focus on interesting work”. It also believes the tool has the potential to lower barriers to entry, enabling more people to explore software development and join the next generation of developers.

GitHub added that the new tool works with a broad set of frameworks and languages, but the technical preview, which is now open to be signed up to, works “especially well” for Python, JavaScript, TypeScript, Ruby and Go.

The company benchmarked the tool against a set of Python functions and blanked out the function bodies and asked GitHub Copilot to fill them in. The model got them right 43% of the time on the first try and 57% of the time when allowed 10 attempts.

If the technical preview is a success, GitHub plans to build a commercial version of GitHub Copilot in the future. Through the preview it aims to learn how people use the tool and what it takes to operate at scale.