Shulou(Shulou.com)11/24 Report--

[new Zhiyuan Guide] AI coding artifact Copilot model upgrade, 80% of the code is automatically generated in 5 years, and the speed of millions of programmers is 55% faster.

Programmer AI artifact has been upgraded again!

Just now, Github announced that the Copilot model has been upgraded and 80% of the code will be automatically generated within 5 years.

Less than two years after its release, GitHub Copilot has written 46% of its code for more than 1 million developers and increased its coding speed by 55%.

What exactly has been improved by this upgrade?

Over the past five years, 80% of the code has been generated automatically through improved models and enhanced context filtering, so developers can now get more tailored suggestions to meet their needs when writing code.

And, draw the point! Both the personal version and the enterprise version can be used without pressure.

According to the official blog, the improved AI model goes beyond the previous Codex model and can provide developers with code advice more quickly.

The new model, developed jointly by OpenAI, Azure AI and GitHub, reduces latency by 13% compared to the previous model.

This means that GitHub Copilot generates code recommendations for developers faster than ever before and promises to significantly improve overall productivity.

At the same time, the new model also has more complex context filtering capabilities, which can take into account the developer's context and usage patterns more widely.

In this way, it can filter tips and code suggestions more intelligently, allowing developers to get suggestions for their needs.

The data show a relative 6% increase in code acceptance, allowing developers to focus on the creative aspects of their work rather than being bothered by tedious coding tasks.

GitHub Copilot is being upgraded to improve the artificial intelligence model and enhance context filtering to provide developers with faster and more tailored code suggestions.

Github is like a second brain that can save you the trouble of remembering.

Github's CEO Thomas Dohmke has also said that 80 per cent of the code will be generated in just five years.

He also says that 40% of the code in the Copilot beta is generated, which increases the speed of developers by 55%.

New generation code Copilot X after Microsoft integrated GPT-4 capabilities into Office 365, GitHub Zeng Guanxuan released a new generation code generation tool Copilot X based on GPT-4.

Specifically, the experiences supported by Copilot X are: Copilot Chat;Copilot for Pull Request;Copilot for Docs;Copilot for CLI.

These new features are driven by OpenAI's GPT-4.

It is worth noting that due to speed delays, the code completion tool is still based on GitHub's Codex model, which is trained on GPT-3.

Prior to this, Microsoft has integrated GPT-4 into search, office, code and other real productivity tools, it is true that it has the power to start the fourth technological revolution.

William Zhu, a senior big data architect, said that Everything powered by AI is no longer far away:

If OpenAI GPT-4 is just a model from 0 to 1, Microsoft has taken a big step forward in its commercialization.

This time, the release of Copilot X directly reduced the dimension to hit the previous generation of Copilot.

For example, GitHub embeds a GPT-4-based chat window in Copilot, focuses on developer scenarios, and integrates on VS Code and Visual Studio.

Copilot can not only identify the code content entered by the developer, display the error message, but also deeply analyze and explain the use of the code block and generate unit tests.

You can even give debug's advice.

In addition, in Copilot, you don't even need a keyboard to write code.

Just sit in front of the computer and say, "Hey, GitHub!" "it can be realized by coding with open mouth.

Currently, GitHub is experimenting with new features of GitHub Copilot Voice, a voice-based interactive system.

In addition to writing code, you can even do it through Copilot Voice: code jump, control IDE, code summary.

Thomas Dohmke, CEO of GitHub, has said that while automatic code completion has greatly increased developer productivity, the new Copilot X can increase developer productivity tenfold.

"the golden age of software development has come to the chat interface. "

55% increase in productivity as early as February, GitHub released major updates to personal and enterprise Copilot.

To put it simply, the upgraded GitHub Copilot will have higher code quality and faster response time.

Since its launch, GitHub Copilot has provided more than 1 million developers with greater productivity, helping them speed up their coding by 55%.

But when it was first launched in June 2022, only 27% of developers chose to use GitHub Copilot-generated code.

Today, that number has risen to 46%. Even in Java, it reached 61%.

Research shows that 90% of developers using GitHub Copilot say they can finish tasks faster, and 73% of them are better able to maintain smoothness and save energy.

At the same time, up to 75% of developers feel more fulfilling when using Copilot and are able to focus on their work.

To achieve this goal, GitHub has made the following key technical improvements:

-upgraded AI Codex model

Upgrading Copilot to a new OpenAI Codex model provides better results for code synthesis.

-better context understanding

The code advice given by GitHub Copilot is improved through a new paradigm called Fill-In-the-Middle (FIM).

This approach takes into account not only the prefix of the code, but also the known code suffix, leaving a gap in the middle for GitHub Copilot to fill.

In this way, Copilot has more context information about the expected code and how it should be consistent with the rest of the program.

-lightweight client model

The GitHub Copilot extension of VS Code has been updated with a lightweight client model, thus improving the overall acceptance of the suggested code.

GitHub Copilot now reduces unnecessary recommendations by 4.5% by using basic information about the user's context (for example, whether the previous recommendation was accepted)

Standard Copilot, Google Colab enlarged recruitment of Microsoft Copilot training upgrade, Google is not to be outdone.

In May, Google announced that Google Colaboratory (Colab) was about to add a new AI coding feature.

On the basis of PaLM 2, after fine-tuning with a large amount of high-quality code data, a new "Vincent code" model Codey was born.

And these new functions of Colab are blessed by Codey.

The Codey code generation model supports more than 20 coding languages, including Go, Google standard SQL, Java, Javascript, Python and Typescript.

Through real-time code completion and generation, Codey can help users complete development work faster and improve the quality of code at the same time.

Most importantly, the model is also specifically optimized for the various functions of Python and Colab.

It can be seen that Google is really careful in order to learn deeply about the experience of developers of applications and Python.

Which do you prefer the coding artifact of Microsoft or Google?

Reference:

Https://the-decoder.com/new-ai-model-boosts-github-copilots-code-generation-capabilities/

Https://web.archive.org/web/20230728192619/https://github.blog/2023-07-28-smarter-more-efficient-coding-github-copilot-goes-beyond-codex-with-improved-ai-model/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.