Pour one into CodeWhisperer, Amazon's AI-powered assisted coding tool. As of today, it's kaput — in a way.
CodeWhisperer is now Q Developer and is part of Amazon's Q family of business-oriented generative AI chatbots, which also extends to the newly announced Q Business. Q Developer, available through AWS, like CodeWhisperer, helps developers with some of the tasks they do in their day-to-day work, such as debugging and upgrading apps, troubleshooting, and running security scans.
In an interview with TechCrunch, Doug Seven, general manager and director of AI developer experience at AWS, hinted that CodeWhisperer is a bit of a failure on branding. Third-party metrics are reflected as wellEven with the free tier, CodeWhisperer struggled to match the momentum of its biggest rival GitHub Copilot, which has more than 1.8 million paid individual users and tens of thousands of enterprise customers. (Poor early impressions certainly didn't help.)
“CodeWhisperer is our starting point [with code generation]But we really wanted to have a brand and a name that fit a broader range of use cases,” Seven said. “You can think of Q Developer as a broader evolution of CodeWhisperer.”
To that end, Q Developer can not only generate code containing SQL, a programming language commonly used to create and manage databases, but also test that code and generate new code conceived from developer queries. We assist in the conversion and implementation of
Similar to Copilot, customers can tweak Q Developer with an internal codebase to improve the relevance of the tool's programming recommendations. (CodeWhisperer, which is now deprecated, also offered this option.) Thanks to a feature called agents, Q Developer also allows you to implement features, document and refactor (i.e. rebuild) your code, and more. tasks can be performed autonomously.
When you make a request to Q Developer, such as “Create an 'Add to Favorites' button in your app,” Q Developer analyzes your app's code, generates new code if necessary, and provides a step-by-step plan. Create and complete testing. Modify your code before implementing the suggested changes. Developers can review and iterate on their plans, connect steps, and apply updates across required files, code blocks, and test suites before Q implements the plan.
“What happens behind the scenes is Q Developer actually launches the development environment and processes the code,” Seven said. “So for feature development, Q Developer takes the entire code repository, creates a branch of that repository, analyzes the repository, performs the requested work, and returns the code changes to the developer. ”
Image credit: Amazon
According to Amazon, the agent can also automate and manage the code upgrade process, and currently offers Java transformations (specifically from Java 8 and 11 to Java version 17 built using Apache Maven). .NET conversion will be available soon. “Q Developer analyzes your code, looks for what needs to be upgraded, makes all the changes, and then hands the code back to the developer to review and commit,” he added. Ta.
To me, agents are a lot like GitHub's Copilot Workspace, which similarly generates and implements plans for software bug fixes and new features. And like Workspace, I'm not entirely convinced that this more autonomous approach will solve the problems surrounding AI-powered coding assistants.
GitClear analyzed more than 150 million lines of code committed to project repositories over the past few years and found that Copilot is pushing more incorrect code into codebases. Elsewhere, security researchers have warned that Copilot and similar tools can amplify existing bugs and security issues in software projects.
This is not surprising. The AI-powered coding assistant is impressive. But they are trained on existing code, and their suggestions reflect patterns of other programmers' work, work that may have serious flaws. The assistant's assumptions lead to bugs that are often difficult to spot, especially when developers who employ large numbers of AI coding assistants don't follow the assistant's decisions.
In less risky areas outside of coding, Q Developer can help businesses manage their cloud infrastructure on AWS. Or at least get the information you need to manage it yourself.
Q Developer can fulfill requests such as “list all Lambda functions” or “list resources that exist in other AWS Regions.” This bot, currently in preview, generates (but cannot execute) AWS Command Line Interface commands and his AWS costs, such as “What were the top 3 most expensive services in Q1?” You can also answer related questions.
Image credit: Amazon
So, how much do these useful generative AIs cost?
Q Developer is available for free in IDEs such as the AWS console, Slack, Visual Studio Code, GitLab Duo, and JetBrains, but with limitations. The free version does not allow for tweaking custom libraries, packages, or APIs, and users choose a data collection scheme by default. You are also subject to monthly limits, including up to 5 agent tasks (e.g. feature implementations) per month and 25 queries per month for AWS account resources. (It's baffling to me that Amazon would limit the questions you can ask about its services, but it's true.)
The premium version of Q Developer, Q Developer Pro, costs $19 per user per month and adds higher usage limits, tools to manage users and policies, single sign-on, and perhaps most importantly, IP coverage.
Image credit: Amazon
The models behind code generation services such as Q Developer are often trained on copyrighted code or under restrictive licenses. Vendors argue that fair use protects them if a model is knowingly or unknowingly developed using copyrighted code, but not everyone agrees. GitHub and OpenAI are being sued in a class action lawsuit alleging that Copilot infringed on copyright by recirculating licensed code snippets without providing credit.
Amazon will defend Q Developer Pro customers against claims that the service infringes third party intellectual property rights, as long as the customers let AWS control the defense and settle it “as AWS deems appropriate.” It says.