80% code generated in seconds! The AI artifact Copilot has been greatly upgraded, and millions of developers will be able to code with their mouths within 5 years.

**Source: **Xinzhiyuanxinzhiyuan‌

**Guide:**AI coding artifact Copilot model upgrade, 80% of the code is automatically generated in 5 years, and the speed of millions of code farmers is increased by 55%.

The Coder AI Artifact has been upgraded again!

Just now, Github officially 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 the code for more than 1 million developers and increased the coding speed by 55%.

What exactly does this upgrade improve?

5 years, 80% of the code is automatically generated

Through the improvement of the model and the enhancement of the context filtering function, developers can now get more tailor-made suggestions to meet their needs when writing code.

And, focus! Both Personal and Enterprise editions can be used stress-free.

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

The new model, developed in collaboration with OpenAI, Azure AI, and GitHub, reduced latency by 13 percent compared to the previous model.

This means that GitHub Copilot can generate code suggestions for developers faster than ever before, which is expected to greatly improve overall productivity.

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

This allows it to filter hints and code suggestions more intelligently, allowing developers to get recommendations specific to their needs.

Data shows a 6% relative increase in code acceptance rates, allowing developers to focus on the creative aspects of their work rather than being bogged down by tedious coding tasks.

GitHub Copilot is getting an upgrade with improved AI models and enhanced contextual filtering to provide developers with faster, more tailored code suggestions.

Github is like a second brain, saving you the trouble of remembering.

Github CEO Thomas Dohmke also said that 80% of the code will be generated in just five years.

He also said that 40% of the code in Copilot beta is generated, which gives developers a 55% increase in velocity.

New Generation Generation Code Copilot X

After Microsoft integrated GPT-4 capabilities into Office 365, GitHub officially announced the release of Copilot X, a new generation of code generation tools based on GPT-4.

Specifically, Copilot X supports the following experiences: Copilot Chat; Copilot for Pull Request; Copilot for Docs; Copilot for CLI.

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

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

Previously, Microsoft has integrated GPT-4 into various real productivity tools such as search, office, and code writing. It is true that it has the power to start the fourth technological revolution.

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

If it is said that OpenAI GPT-4 is just a model that has completed from 0 to 1, Microsoft has promoted its commercialization a big step forward. This time, the release of Copilot X directly hits the previous generation of Copilot with dimensionality reduction.

For example, GitHub embeds a GPT-4-based chat window in Copilot, focusing on developer scenarios, and integrating it with VS Code and Visual Studio.

Copilot can not only identify the code content entered by the developer and display the error message, but also conduct in-depth analysis and explanation of the purpose of the code block, and generate unit tests.

You can even give debugging suggestions.

Plus, in Copilot, you don't even need a keyboard to write code anymore.

Just sit in front of the computer, say "Hey, GitHub!", and start coding with your mouth.

Currently, GitHub is experimenting with a new feature called GitHub Copilot Voice, a voice-based interaction system.

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

GitHub CEO Thomas Dohmke once said that although automatic code completion has greatly improved the productivity of developers, the new Copilot X can increase the productivity of developers by 10 times.

"The golden age of software development has come to chat interfaces."

Work efficiency increased by 55%

Back in February, GitHub released a major update to the personal and enterprise versions of Copilot.

Simply put, the upgraded GitHub Copilot will have higher code quality and faster response speed.

Since its release, GitHub Copilot has empowered more than one million developers to be more productive, helping them code 55% faster.

But only 27% of developers will choose to use the code generated by GitHub Copilot when it first launches in June 2022.

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

Research shows that 90% of developers who use GitHub Copilot say they can complete tasks faster, and 73% of them are better able to stay in the flow and save energy.

Meanwhile, a whopping 75% of developers feel more fulfilled and able to focus on their work when using Copilot.

In order to achieve this goal, GitHub has made the following key technical improvements:

• Upgraded AI Codex model

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

• better contextual understanding

Code suggestions given by GitHub Copilot are improved through a new paradigm called Fill-In-the-Middle (FIM).

This approach not only takes into account code prefixes, but also leverages known code suffixes and leaves gaps in the middle for GitHub Copilot to fill in.

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

• Lightweight client model

Updated the GitHub Copilot extension for VS Code with a lightweight client model, improving the overall acceptance rate for suggested code.

GitHub Copilot now reduces unnecessary suggestions by 4.5% by using basic information about user context (e.g. whether the last suggestion was accepted)

Compared to Copilot, Google Colab zooms in

After Microsoft Copilot practiced and upgraded, Google is not to be outdone.

In May, Google announced that Google Colaboratory (Colab) will soon add a new AI coding function.

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

These new features of Colab are supported by Codey.

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

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

Most importantly, this model is also specially optimized for various functions of Python and Colab.

It can be seen that Google really cares about the experience of developers of deep learning applications and Python.

Which coding artifact do you prefer between Microsoft and Google?

References: