GitHub also said in a statement that the AI-powered tool, built using OpenAI’s Codex model, currently writes 46% of the code on the platform and has helped developers code up to 55% faster. According to him, “A developer can get in-depth analysis and explanations of what code blocks are intended to do, generate unit tests, and even get proposed fixes to bugs.”
It removes the silos in companies and I think that’s where we will unlock potentially even more than 10x in productivity.“This does far more than suggest code,” explained Dohmke, stating that GitHub Copilot Chat recognises what code a developer has typed, what error messages are shown and it’s deeply embedded into the integrated development environment (IDE), ready to accept commands. “Once you have a Copilot in every single application - whether that’s in Outlook, Word or your code editor, you can leverage the things you wrote in those environments, but you can also leverage the information from other workers in your organization, whether that’s other developers or maybe the marketing person, who now has an easier time writing the announcement blog post, because they have seen - in their Copilot - the release notes that the developer has put into their git repo. That’s not quite ready for public consumption, though. What’s maybe more interesting, though, is that GitHub plans to allow organizations to use this tool on top of their repositories and internal documentation. GitHub is starting with documentation for React and from the Azure Docs and Mozilla’s MDN and developers will be able to query those docs in a ChatGPT-like interface. But ChatGPT’s cutoff date in 2021 means this information isn’t always up to date, so GitHub is augmenting it with more recent data (similar to what Microsoft’s Bing is doing for search with its Prometheus model).
Also new is Copilot for docs, which enriches ChatGPT’s existing database, which you can already query about framework and language docs, for example.
0 kommentar(er)
