Go to our on-demand library to view VB Transform 2023 sessions. Register here
by GitHub latest breakthrough in Generative AI and GPT-4, Co-Pilot X, is expanding its reach to corporate businesses and organizations. Today, the company announced the limited public beta of GitHub Copilot Chat. With this, GitHub aims to integrate a context-aware conversational assistant directly into integrated development environments (IDEs) such as Microsoft Visual Studio and VS Code.
According to GitHub, developers will be able to effortlessly tackle complex tasks through simple prompts using Copilot Chat. The company says this will allow every member of the development team, regardless of experience level, to build complete applications or debug extensive codebases in minutes instead of days.
“Unlike a generic generative AI chat assistant, Copilot Chat is built specifically for developer scenarios and is contextually aware of the code a developer has typed and any error messages they see because it’s right there with them in their code editor /IDE, where they spend most of their time programming,” Mario Rodriguez, vice president of product management at GitHub, told VentureBeat.
Rodriguez said the company’s latest offering is an AI duo programmer, explicitly designed to assist developers with numerous tasks, such as launching a file in an unfamiliar coding language or framework, autocomplete boilerplate code and debugging and writing unit tests.
GitHub says the new offering will democratize software development, improving the productivity and satisfaction of developer teams.
“Most of the AI developer tools are either experimental or have yet to be trialled at scale, while Copilot Chat is built on GitHub Copilot, which is already trusted by over 1 million developers,” said Rodriguez of GitHub . “So we benefited from being first to market, defining how AI can best improve developer workflow, and refining GitHub Copilot based on feedback from such a large user base.”
Provide developer assistance through contextual understanding
GitHub said that Copilot Chat surpasses the functionality of a typical chat window. It can understand code written by a developer and interpret the error messages that are displayed. The company says that unlike generic generative AI chat assistants, Copilot Chat demonstrates contextual awareness, integrating effective concepts for generic AI and tailoring them to specific developer environments.
“Copilot Chat is contextually aware of what a developer is trying to do at any given moment. That context allows it to provide guidance specifically tailored to the user rather than offering general suggestions that may not apply to that scenario,” Rodriguez told VentureBeat.
GitHub says developers previously lacked an easy way to investigate or get additional context. With Copilot Chat, they can access instant, context-specific support right in their Editor/IDE.
“You can ask Copilot to propose a fix for bugs in your code. By looking at your comment and comparing it to the code, Copilot will not only acknowledge errors and provide context on what went wrong, but will also propose fixes that will resolve the issues,” Rodriguez said.
The contextual approach of the AI model addresses the challenge of maintaining developer workflow amid the increasing complexity of programming over the past two decades. Factors contributing to this complexity include the proliferation of languages, cloud computing, programming frameworks, and different services.
For example, developers don’t need to use a regular expression translator when faced with poorly documented regular expressions. Instead, they can simply highlight the code and request explanations from the Copilot chat.
Beyond understanding
In addition to understanding the code, developers can improve their code by instructing Copilot Chat to “improve code readability”, “add more comments”, or “separate validation function”.
“Users can ask Copilot chat for help with coding challenges. If Copilot Chat doesn’t fully answer your question with its first response, you can continue to ask follow-up questions, request clarifications, and more,” Rodriguez said. “This conversational element makes Copilot Chat so powerful—it’s not a tool complete; it’s a conversational assistant that stays with you throughout the entire coding process.”
The company claims incredible productivity gains with GitHub Copilot. In a controlled study, GitHub found that developers completed tasks 55% faster using GitHub Copilot. Early research indicates that an average of 46% of code across all programming languages is built with GitHub Copilot, a number that rises to 61% among Java developers.
Security control
Rodriguez said users can ask Copilot Chat to review their code within the IDE itself. During this review process, Copilot Chat may identify potential security issues and offer suggestions for correction.
“What makes Copilot Chat especially unique for this scenario is that the results are tailored to the user’s code, whereas if a developer had searched Stack Overflow or Google, they could have come across dozens of variants, templates and flavors for solutions a bug, and one relevant to the user might not even be one of them,” Rodriguez explained. “Ultimately, this feature can reduce the number of vulnerabilities found in security scans.”
GitHub said developers can converse with Copilot Chat using natural language, just like a human programmer, allowing for discussion of complex concepts. The company says this approach surpasses conventional methods of searching and reading documentation.
“Instead of interrupting what they’re doing to look for functionality in a code snippet, they can simply ask Copilot Chat and get an answer right in the IDE. It saves time and makes coding more interactive and engaging,” Rodriguez told VentureBeat. “We also believe Copilot Chat will lower barriers to entry and help novice programmers improve their skills faster.”
VentureBeat’s mission it is to be a digital city square for technical decision makers to gain insights into transformative business technology and transactions. Discover our Briefings.
