Amazon is proud to announce CodeWhisperer at the re: Mars conference on page 624 as a GitHub Copilot-like AI programming aid that helps developers easily make up the code they need. Initially, Amazon conducted training for the Java, JavaScript, and Python programming languages based on billions of lines of publicly accessible source code, its own codebase, and sample documentation on public forums.
CodeWhisperer is now available in preview as part of the AWS IDE Toolkit, and interested developers can immediately experience it through their preferred integrated development environment.
These include VS Code, IntelliJ IDEA, PyCharm, WebStorm, and Amazon’s own AWS Cloud 9, as well as support for the AWS Lambda console soon.
Before today’s official announcement, Vasi Philomin, vice president of Amazon’s AI services, stressed that the company is not simply creating a replica of GitHub Copilot.
In fact, CodeWhisperer was implemented on years of technology accumulation such as the company’s AI code reviewer, CodeGuru performance analyzer, and DevOps Guru tools for slot operation issues.
In order to maintain a high degree of confidentiality, the project was initially known only to a small number of internal testers. Now, however, interested developers can use CodeWhisperer to easily check code comments and even recommend variable names based on their personal programming style.
It’s worth mentioning that CodeWhisperer also does something different from GitHub Copilot. On the one hand, although much of the code generated by the system is novel, it emphasizes the original license every time a fragment similar to an existing fragment is generated during training.
By putting the choice in the hands of developers, CodeWhisperer helps mitigate potential copyright issues that may be encountered when using such tools.
Second, Philomin highlights the security features of CodeWhisperer. Thanks to Amazon’s own large code base and extensive problem reporting experience, this AI programming aid scans code for potential security issues.
Finally, the CodeWhisperer team is working to ensure that code without any bias is generated (by deploying filters), resulting in a fairly neutral AI-assisted programming experience for developers.
