OpenAI’s new neural network Codex writes software in response to text prompts


OpenAI LLC today released the private beta program for an artificial intelligence system called Codex that can write software and perform data science tasks in response to natural language prompts.

Codex will be available to participants in the private beta program through an application programming interface. The API will allow participants to build their own custom software services on top of the AI ​​system.

Codex is an improved version of an AI of the same name, for which OpenAI was developed GitHub co-pilot, a developer productivity tool launched last month. Both algorithms are based on GPT-3, a groundbreaking neural network that OpenAI researchers first described in detail last year. GPT-3 can write essays on business topics, create web pages, and perform other tasks based on simple instructions given by the user in natural language.

OpenAI’s Codex system takes the same approach to coding. A developer can provide a natural language prompt, such as: B. “Create a user interface with 10 buttons” and Codex will automatically generate source code that will perform the requested task. It is also capable of writing much more sophisticated software.

In a project detailed by OpenAI, researchers had Codex analyze weather data from the US National Oceanic and Atmospheric Administration to create a graph of daily peak temperatures in San Francisco. In another experiment, OpenAI staff used Codex to create a video game. The AI ​​system even managed to rewrite code written in a programming language, Python, into the Ruby language.

Codex supports more than a dozen programming languages ​​at startup, although it does Python best. Python is widely used for data science tasks such as creating visualizations of business data. The language is also popular with AI developers.

Codex “has 14 KB of memory for Python code, compared to GPT-3, which is only 4 KB – so it can take more than three times as much contextual information into account when performing any task,” said the OpenAI researchers.

“Once a programmer knows what to build, the process of writing code can be thought of as (1) breaking a problem down into simpler problems, and (2) breaking those simple problems onto existing code (libraries, APIs, or functions) already exist, ”added the researchers. “The latter activity is probably the least fun to program (and has the highest barrier to entry), and this is where OpenAI Codex excels most.”

Another version of Codex is available as part of GitHub Copilot, a developer tool that Microsoft subsidiary GitHub launched last month. It’s a sophisticated autocomplete engine that allows programmers to start writing a snippet of code and, in some cases, have the next few lines autocomplete. However, GitHub Copilot doesn’t have a natural language interface like the new edition of Codex that OpenAI debuted today.

Both neural networks are particularly adept at integrating external software components into a developer’s code. A significant part of the job of software teams is integrating external components such as databases into their applications.

The task is relatively simple in many cases, but it requires developers to spend a lot of time reading technical manuals. By automating the process, GitHub Copilot can save developers the hassle of digging through guides and, in some cases, potentially saving them hours of work.

GitHub parent Microsoft is a major supporter of OpenAI and invested $ 1 billion in the AI ​​research lab two years ago to support its work. The technology giant is also a leading provider of developer tools. In addition to owning GitHub, the industry’s leading platform for hosting code, Microsoft has several other products that are widely used in software development projects, including the Visual Studio code editor.

Microsoft can finally make Codex available as a cloud service for corporate customers. The company previously purchased an exclusive license for GPT-3, the AI ​​on which Codex is based, from OpenAI and plans make the technology available through Azure.

Tools that can improve developer productivity are big business. Sourcegraph Inc., a startup that saves software teams time by helping developers explore application code faster, recently Raised $ 125 million from investors. A theoretical future Codex-based Azure service that streamlines not just a number of development tasks, but many and across multiple programming languages, may have the potential to generate significant revenue for Microsoft.

This is especially true as OpenAI continues to improve Codex by increasing the number of tasks the AI ​​can automate. To support OpenAI’s research, Microsoft has a dedicated supercomputer created for the lab and made available in Azure. The system includes no fewer than 10,000 graphics processors and 285,000 processor cores with which the laboratory’s scientists train new AI models.

Image: OpenAI

Show your support for our mission by joining our Cube Club and Cube Event Community of Experts. Join the community that includes Amazon Web Services and soon Andy Jassy, ​​CEO of, Michael Dell, founder and CEO of Dell Technologies, Pat Gelsinger, CEO of Intel and many more luminaries and experts.

join our community

On June 16, we are holding our second cloud startup showcase. Click here to take part in the free and open Event Startup Showcase.

“TheCUBE is part of re: Invent, you know, you are really part of the event and we are very happy that you are coming here and I know that people also appreciate the content you create.” – Andy Iassy

We really want to hear from you. Thank you for taking the time to read this post. We look forward to your visit at the event and in the CUBE Club.


Leave A Reply