In late June 2021, Microsoft, owner of the collaborative web platform GitHub, and Open AI launched a technical preview of GitHub Copilot, an AI-powered completion tool designed to help developers code faster. It is also available to them as a subscription from June 21, but is free for students and managers of popular open source projects. GitHub surveyed more than 2,000 developers who participated in a technical review, and a recently published study found that Copilot increases their productivity.
Completion tools they are designed to enable developers to increase productivity in increasingly complex activities. GitHub Copilot, an extension of Visual Studio Code, runs OpenAI Codex, an AI system created by OpenAI, that performs better than GPT-3 in generating code, in part because it is trained on a dataset that includes a much higher concentration of public source code. GitHub Copilot works with a wide range of frameworks and languages, but works especially well for Python, JavaScript, TypeScript, Ruby, and Go. It offers suggestions that the developer is free to accept or reject.
The launch of Copilot’s technical preview caused a lot of reaction in the developer community, and especially in the La community Free Software Foundation regarding copyright, AI-generated code ownership issues, and legal implications for GitHub authors. He then launched a call for white papers on copilot, copyright, machine learning and free software, selected and published 5 on 24 February.
GitHub’s study of developer productivity using Copilot
Last April, GitHub conducted the first study with 24 users to understand how developers use and experience Copilot, finding that while it did not reduce task time or increase success rates, most participants preferred to use Copilot in everyday development tasks because it often provided a useful starting point and saved the effort of searching online. However, participants had difficulty understanding, editing, and correcting errors in the code snippets generated by Copilot, which significantly hindered their performance in solving the tasks.
For this new study, research and engineering teams teamed up to combine data from a qualitative survey of more than 2,000 US-based developers with anonymized data to determine whether developers believe they are more productive with GitHub Copilot and whether the data proves they are. .
Using GitHub Copilot is associated with improved developer productivity
Developers who report the greatest productivity gains with GitHub Copilot are those who have accepted the largest number of published code proposals:
- Those who accepted 29.73% of the suggestions reported a “large” increase in productivity;
- Those who declared a “high” increase in productivity accepted 27.77% of the proposal;
- Users who reported an “average” increase in productivity accepted 26.94% of suggestions;
- Those who reported “modest” gains accepted 23.16% of the proposals.
The study also finds that adoption rates vary by language, which didn’t surprise GitHub authors, with Javascript and Python benefiting from the highest adoption rates. The study also revealed that developers were primarily looking for a suitable starting point.