Based on GPT-3, they have developed a tool that can turn your commands into lines of code

AI research company OpenAI has just launched a new machine learning tool that can convert English sentences into lines of code. This tool is called Codex and it is designed to speed up the work of professional programmers as well as beginners.

In the tests, Codex was able to create basic web pages and simple games based on English commands. In addition, it can also translate code from one programming language to another as well as solve data science queries.

Just make requests like “create a website with a menu on the side and a title on top” and the Codex will translate it into code. Of course, the Codex hasn’t evolved to the point of infallibility, and you’ll have to have a bit of patience at first to use it.

Previously, OpenAI used a version of Codex to build a tool called Copilot for Microsoft’s Github. Copilot is like Gmail’s automatic sentence completion tool when it provides code suggestions so programmers can get the job done quickly. However, the new Codex has many improvements and is more flexible than Copilot when it can not only perfect the code but also generate the code itself.

The Codex is built on top of GPT-3, OpenAI’s advanced language generation model. GPT-3 is trained on the massive data of the internet, so it can generate and parse written words in impressive ways. One of the applications of GPT-3 is writing code.

However, Codex is an upgraded version of GPT-3 when specially trained based on open source repositories pulled from the web.

The current Codex can create basic websites and simple games. However, to give the correct instructions for this tool, you need to know a little bit about what you want to do, think a little deeper.

The Codex developers insist that this tool is not a creation god that can read your brain and turn every statement into perfect lines of code. When working with it, you will have to be patient and it will take some trial and error.

OpenAI is very optimistic about the potential of Codex. They argue that the tool could solve America’s programmer shortage and be the next step in the evolution of code.

What can Codex do? Is it a programmer’s threat?

Over the summer, OpenAI engineers were excited to test the capabilities of the Codex. However, as they experimented more and more deeply, they realized that the software would not replace the programmers.

The Codex can create programs in 12 different computer languages ​​and can even translate between them. However, it often makes mistakes, and despite its impressive skill, it cannot reason like a human. It can recognize or imitate what it has seen in the past, but it is not intelligent enough to think for itself.

Sometimes, programs created by Codex fail to run. Or, the Codex writes programs with vulnerabilities. Moreover, there are times when the Codex creates programs that are not meant for you. Overall, according to OpenAI statistics, Codex only generates correct code about 37% of the time.

In other words, the Codex is only really useful to an experienced programmer. It can help programmers do their day-to-day tasks a lot faster. It can help them find the building blocks they need or direct them towards new ideas.

In short: “AI won’t take anyone’s work. Instead, it will help us solve the hard, boring things”, said Greg Brockman, chief technology officer of OpenAI.