OpenAI is making an algorithm that can take written commands as input and convert them into chunks of code. OpenAI calling this algorithm “Codex.”
Codex takes written instructions and turns them into usable code. This way, Codex can help to make programming work simple and fast and can help beginners to learn to program.
While showing the demo of Codex, OpenAI shows that how Codex can build simple websites and rudimentary games just by using natural (only English) language. Codex can also convert the code of a programming language into another programming language and understand data science queries. For example, users types, “create a webpage with the title at top and menu at left,” and Codex translates this into code and creates a webpage the same as required. This process is quite slow, yet this can prove to gigantic step towards accessible and faster coding.
“We see this as a tool to multiply programmers. Programming has two parts to it: you have ‘think hard about a problem and try to understand it,’ and ‘map those small pieces to existing code, whether it’s a library, a function, or an API.’ It takes people who are already programmers and removes the drudge work.”
-OpenAI’s CTO and co-founder Greg Brockman to The Verge.
This is not the first time when OpenAI created such a code “helping” algorithm. OpenAI created an earlier version of Codex and used that to create a tool “Copilot” for GitHub – a Microsoft-owned code repository. Copilot was a kind of autocomplete tool, which suggests coders code write next by using AI. Newer Codex is much advanced and not autocomplete but auto-create.
OpenAI builds codex upon GPT-3 – OpenAI’s language generation model, GPT-3 model was trained on an enormous amount of words from the Internet. But, Codex has an improvement that it was trained on open-source code repositories scraped from the Internet.
Codex’s creator’s company said Codex will certainly bring more value and investment but OpenAI will not use Codex to develop its own tools but for core improvement. Brockman said: “We realized if we pursued any one of those, we would cut off any of our other routes. You can choose as a startup to be best at one thing. And for us, there’s no question that that’s making better versions of all these models.”
In this Blog
Codex and the future of Programming
OpenAI is much positive about how Codex will make programming easy and efficient. Brockman said Codex has the potential to solve the shortage of programmers in the US and Codex lead Wojciech Zaremba said Codex can be the next step towards the historic evolution of coding.
Zaremba further said: “What is happening with Codex has happened before a few times. These programming languages, resembled English, using vocabulary like ‘print’ or ‘exit’ and so more people became able to program. Each of these stages represents programming languages becoming more high level. And we think Codex is bringing computers closer to humans, letting them speak English rather than machine code.”
Codex can understand and speak more than a dozen languages like Ruby, PHP, Pearl, Swift, and Typescript, but it is best in Python.
Not just coding, but Codex can also control other computer programs. In a demo, OpenAI showed that how Codex can create a voice interface for MS word. MS word has its own API and Codex can give him instructions generated from users’ voice commands.
In this way, Codex can prove to be an interface between the user and computer programs. According to OpenAI, they also successfully tested Codex with Spotify and Google calendar. Brockman said: “They’re [Microsoft] very excited about the model in general and you should expect to see lots of Codex applications be created.”