Is an IDE with AI the way of the future for making software? GitHub is throwing out the idea.
GitHub launched Copilot Workspace before its annual GitHub Universe conference in San Francisco early this fall. Copilot Workspace is a development environment that uses “Copilot-powered agents” to help developers use natural language to come up with ideas, plan, build, test, and run code.
Jonathan Carter, who is in charge of GitHub’s software research and development team called GitHub Next, says that Workspace is like a more general version of Copilot, GitHub’s AI-powered coding assistant. It builds on new features like Copilot Chat, which lets developers ask questions about code in natural language.
Carter said, “Our research showed that for many tasks, the hardest part for developers was just getting started, especially when it came to knowing how to approach a [coding] problem, which files to edit, and how to think about different solutions and the pros and cons of each.” “That’s why we set out to make an AI assistant that could meet developers at the start of an idea or task, lower the amount of energy needed to get going, and then work with them to make the necessary changes across the whole corebase.”
Copilot had over 1.8 million paying individual customers and 50,000 business users as of the last count. But Carter sees a much bigger base that will be drawn in by feature additions that appeal to a wide range of people, like Workspace.
“Our goal is to give developers more power every day through a ‘thought partnership’ with AI, since they spend a lot of time working on [coding issues],” Carter said. “You can think of Copilot Workspace as an experience and development environment that works with other tools and workflows and makes it easier for developers to do certain tasks…” There is a lot of value that can be provided in an AI-native developer environment that isn’t limited by how things are done now.
There is definitely pressure inside Copilot to make money
A story in the Wall Street Journal says that Copilot loses $20 a month on average per user. Some customers cost GitHub as much as $80 a month. And the number of services that compete with them keeps going up. There’s Amazon CodeWhisperer, which the company gave away for free to coders at the end of 2018. There are also new businesses, such as Magic, Tabnine, Codegen, and Laredo.
Working with OpenAI’s GPT-4 Turbo model, Workspace can create a plan to (attempt to) fix a bug or add a new feature based on a GitHub repository or a specific bug within a repository. It does this by looking at the repository’s comments, issue answers, and larger codebase. Developers are given suggested code for fixing bugs or adding new features, along with a list of the tools they need to test and confirm that code. They are also given the option to change, save, refactor, or undo the code.
The suggested code can be run right in Workspace and sent to other team members through an outside link. When those team members get into Workspace, they can make changes to the code however they want.
Workspace can be opened from the new “Open in Workspace” button to the left of issues and pull requests in GitHub repos. This may be the easiest way to get to Workspace. When you click on it, a field appears where you can describe the software engineering job in plain English, such as “Add documentation for the changes in this pull request.” When you send it, it is added to a list of “sessions” in the new Workspace view.
Workspace handles requests in a planned, step-by-step way, first making a blueprint, then a plan, and finally putting that plan into action. Developers can look more closely at any of these steps to see the suggested code and changes. If needed, they can remove, re-run, or change the order of the steps.
“Every developer will tell you that they get stuck on a new project because they don’t know where to start,” Carter said. “Copilot Workspace takes that load off developers and gives them a plan to start building on.”
On Monday, Workspace goes into technical preview. It is optimized for a number of platforms, including mobile ones.
It’s important to note that Workspace isn’t covered by GitHub’s IP protection policy because it’s still in preview. This policy says that GitHub will help pay for customers’ legal fees if third parties claim that the AI-generated code they’re using violates their IP. It is well known thatgenerative AI models repeat their training datasets, and GPT-4 Turbo was learned in part on code that was protected by intellectual property laws.
It’s up to GitHub to decide how to make Workspace a full product. In the meantime, it says it will use the preview to “learn more about the value it delivers and how developers use it.”
I think the more important question is: Will Workspace fix the problems that Copilot and other AI-powered writing tools have with their very existence?
GitClear, the company that made the same-named code analysis tool, looked at more than 150 million lines of code that had been committed to project repos over the past few years. They found that Copilot was causing more mistaken code to be pushed to codebases and more code to be added again instead of being reused and streamlined, which was making things harder for people who maintain code.
In other places, security experts have said that Copilot and other tools like it can make bugs and security problems in software projects even worse. Also, experts at Stanford have found that developers who use AI-powered coding assistants to help them write code tend to write less secure code. GitHub told me that it has an AI-based vulnerability prevention system that tries to stop unsafe code and an optional code copying filter that looks for copies of public code.
But developers aren’t afraid of AI
44% of coders polled on StackOverflow in June 2023 said they use AI tools in their work now, and 26% said they plan to start adding them soon. By 2028, Gartner says that 75% of corporate software engineers will use AI code assistants.
Workspace might be able to help clean up some of the mess that AI-generated code causes by putting more emphasis on human review. We’ll know soon enough, once workers get their hands on Workspace.
Also Read: The Company Really Wants to Talk About Copilot
Carter said, “Our main goal with Copilot Workspace is to use AI to make things simpler so developers can be more creative and explore more freely.” “We really believe that HR and AI working together will always be better than either one alone, and that’s what Copilot Workspace is based on.”
What do you say about this story? Visit Parhlo World For more.
