Last June, Microsoft-owned GitHub and OpenAI launched Copilot, a service that provides suggestions for whole lines of code inside development environments like Microsoft Visual Studio. Available as a downloadable extension, Copilot is powered by an AI model called Codex that’s trained on billions of lines of public code to suggest additional lines of code and functions given the context of existing code. Copilot can also surface an approach or solution in response to a description of what a developer wants to accomplish (eg, “Say hello world”), drawing on its knowledge base and current context.
Copilot was previously only available in technical preview. But after signaling that the tool would reach general availability this summer, GitHub today announced that Copilot is now available to all developers. As previously detailed, it’ll be free for students as well as “verified” open source contributors – starting with roughly 60,000 developers selected from the community and students in the GitHub Education program.
GitHub says that 1.2 million people signed up during the preview period. Copilot is now suggesting 40% of newly written code, according to the company – up from 35% earlier this year.
“Over the past year, we’ve continued to iterate and test workflows to help drive the‘ magic ’of Copilot, ”Ryan J. Salva, VP of product at GitHub, told TechCrunch via email. “We not only used the preview to learn how people use GitHub Copilot but also to scale the service safely. ”
Copilot extensions are available for Noevim and JetBrains in addition to Visual Studio Code, or in the cloud on GitHub Codespaces.
One new feature coinciding with the general release of Copilot is Copilot Explain, which translates code into natural language descriptions. Described as a research project, the goal is to help news developers or those working with an unfamiliar codebase.
“While it’s clear that Copilot helps developers complete tasks faster, we’re continuing to explore updates that go beyond that by helping developers stay in the flow, focus on more satisfying work, and conserve mental energy even as they save time, ”Salva said. “As an example of the impact we’ve observed, it’s worth sharing early results from a study we are conducting. In the experiment, we are asking developers to write an HTTP server – half using Copilot and half without. Preliminary data suggests that developers are not only more likely to complete their task when using Copilotbut they also do it in roughly half the time. ”
Owing to the complicated nature of AI models, Copilot remains an imperfect system. GitHub said that it’s implemented filters to block emails when shown in standard formats, and offensive words, and that it’s in the process of building a filter to help detect and suppress code that’s repeated from public repositories. But the company acknowledges that Copilot can produce insecure coding patterns, bugs and references to outdated APIs, or idioms reflecting the less-than-perfect code in its training data.
“This is just the beginning of AI-powered development tools, so it’ll be exciting to see how developers use Copilot over the next few months and years from now – and in tandem, how we advance the product,” Salva continued.