Until recently, coding involved repetitive tasks and required knowing many tiny details. These aspects of coding detract from the creative work developers really enjoy and slow developers down.
Now, AI technology promises to eliminate much of the repetitive work, and developers are no longer left behind having to scour the web for those tiny details to get things done.
This technique works like autocomplete in word processing, but writes code instead of plain language and completes the entire function at once.
AI help algorithm, boilerplate code
Manual in latest product Smart-driven is Github’s Copilot, an AI-powered pair programming tool available to all developers for $10 per month or $100 per year.
The company claims that Copilot can suggest complete methods, boilerplate code, entire unit tests, and even complex algorithms.
“With AI-driven coding techniques like Copilot, it’s very easy to introduce as it works as before, but with greater speed and satisfaction,” explains Oege De Moor, VP of developer GitHub Next “It really helps to be explicit in your instructions to AI.
He explained that during the Copilot tech preview, GitHub heard from users that they wrote better and more accurate explanations in code comments because AI gave them better Recommended.
“Users also wrote more tests because Copilot encouraged developers to focus on the creative part of making good tests,” explains De Moor. “So, these users felt like they wrote Better code, hand in hand with Copilot.
It is of course important to make users aware of the limitations of the technology, he added.
“Like all code, suggestions from AI assistants like Copilot need to go through Carefully tested, reviewed and reviewed,” he said. “We are also constantly working to improve the quality of the recommendations made by the AI. “
GitHub Copilot is built using Codex – a descendant of GPT-3 – which is code and natural language trained on publicly available source code.
“Because it’s trained on both source code and natural language, you can write comments in English, and Codex will suggest the code behind,” De Moor explained. “In fact, it can even write an entire function or class, just describe it in English. ”
Future AI capabilities can aid debugging
Tabnine CEO Official Dror Weiss said that in the future, AI assistants will be able to review code for developers, automatically create tests, assist with debugging, and intelligently automate maintenance operations on the system.
“Ultimately, everything that can be automated Activities will be automated,” he said.
In his view, a key characteristic of an organization is the ability to integrate specific best practices and code patterns for projects and organizations.
“Using this customized AI, organizations will benefit not only from acceleration, but also from better code consistency and quality,” he explained. “Another benefit is that developers spend less time onboarding new The time needed to increase productivity while on a project. “
A major benefit of AI-assisted coding tools is context-aware code completion.
For example, Microsoft’s Visual Studio IntelliCode is a set of AI-assisted features that enable developers to Humans are able to efficiently complete code with features such as parameter completion, code formatting, and style rule referencing.
IntelliCode is trained on code from thousands of highly rated open source projects on GitHub, it Use the context in the current code to make relevant suggestions.
Since the introduction of IntelliCode, Microsoft has made updates such as whole-line code completion and refactoring as well as enhancing the repetitive editing experience to save developers time Suggestions.
Roadmap for AI Coding Assist is key
For organizations planning to implement a strategy involving AI coding assistants, Weiss said a roadmap is key.
“Organizations need to think strategically and have a vision for how they want to leverage AI, even in the marketplace. Any product that doesn’t have some basic functionality yet,” he said.
He explained that the logical first step in implementing AI assistance is to identify a specific set of developers and have them Using artificial intelligence based on pre-trained models that learn code patterns from publicly available code.
After successful implementation, organizations can start rolling out to other groups. At the same time, they can The code creates custom AI models to tailor the AI assistance to your needs.
De Moor also noted that developers spend most of their time on other tasks that will soon benefit as well For AI-assisted.
Examples of these other tasks AI-assisted mature tasks are code review, testing, and refactoring.
“This changes the developer’s job ? Of course, but for the better,” De Moore said. “I don’t foresee a future where Copilot will produce anything useful without human input, but I do see unchecked human creativity, no longer being swayed by irrelevant Bound by the details.
He said programming is now about design (breaking a big problem into small problems) and then specifying what the small pieces should do – the AI will fill in the details.
Weiss added that with every company becoming a “software company,” software development is the most strategic and resource-constrained activity for every organization.
“Companies are starting to meet constraints on how many developers they can get, and making smaller teams more efficient is paramount — even more so in a downturn, teams can be understaffed, “He says. “We believe that AI is the most effective way to increase developer and team productivity, and will be the natural next step for every organization that adopts a fundamental DevOps and CI platform.”
What to read next:
Finding Coding Quality
Modern Application Development: An Enterprise Guide
Can AI lead low-code/no-code app development?
Big shift in software development still requires hardware