OpenAI and GitHub Unveil New Copilot AI Assistant for Coding
A new virtual assistant created by OpenAI and GitHub will suggest code to software developers as they work. The new GitHub Copilot tool leverages an improved version of OpenAI’s popular GPT-3 language model called Codex to teach the AI how to collaborate in a coding project like a human partner.
GitHub Copilot takes the concept of natural language processing and applies it to programming languages. The idea is to imitate a “pair programmer,” when two developers simultaneously work on a coding project and comment and annotate each other’s work along the way. The AI theoretically takes the junior partner role in the endeavor, making its name entirely apropos. Copilot relies on OpenAI’s Codex model to understand what the programmer is doing and come up with suggestions. Like GPT-3, Codex is built on an enormous collection of data to teach an AI how to suggest a line or more of code. The AI learns from what suggestions the human user accepts or rejects, honing its understanding and ideally leading to better code ideas.
“GitHub Copilot draws context from the code you’re working on, suggesting whole lines or entire functions,” GitHub CEO Nat Friedman explained in a blog post about the new tool. “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.”
“GitHub Copilot is powered by OpenAI Codex,” Friedman wrote. “OpenAI Codex has broad knowledge of how people use code and is significantly more capable than GPT-3 in code generation, in part, because it was trained on a data set that includes a much larger concentration of public source code.”