AI turns plain English into software

There are two key considerations when it comes to coding, Greg Brockman, chief technology officer and co-founder of AI research firm OpenAI, told The Verge.

The first part is thinking about the problem, Brockman said, and really understanding it. The second part is to figure out how to solve this problem, using code.

It’s this second aspect that OpenAI’s new system, called Codex, hopes to make easier, faster and more accessible.

The codex can go from text to code, take commands written in plain English and bring them to life.

The codex can go from text to code, take commands written in plain English and bring them to life. Type “build a website with a menu on the left side and a taskbar on the right”, to modify The Verge’s example, and Codex will build it, without you having to type any code. (Now how Good it creates it for you may take some tweaking.)

Figuring out how to solve your problem with the proper code language is “probably the least fun part of programming,” OpenAI wrote on its blog announcing Codex; it is also the highest barrier to entry.

Being able to go from text to code can not only make life easier for programmers, but also open the door to people who don’t know the crucial language of the field: software.

From text to code: The language is kind of a strong point for OpenAI. The company has built powerful algorithms for natural language processing (NLP) – it’s the kind of deep learning AI that allows Siri to understand you.

Their most robust AI language generator to date, GPT-3, was a phenomenon in AI circles when it debuted last summer; Using essentially all written English on the Internet as learning material, GPT-3 is capable of producing amazingly written Human.

According to OpenAI, Codex is a descendant of GPT-3, except it’s not built on the language alone, but also on the huge repositories of open source computer code on the internet – innumerable lines of code.

Armed with this knowledge base, Codex can move from text to code in a conversation-like fashion. WIRED’s Steven Levy got to see the AI ​​in action, as Brockman and fellow co-founder Wojciech Zaremba performed a live demo in San Francisco.

“It… removes the heavy work.”

Greg Brockman

First, they asked Codex to create a website for them, full of text and images, and put it online.

“Then, using the casual language one might use in conversation, they created a simple game by grabbing images of helicopters from the web and having them fly across the screen and blast enemies” , wrote Levy.

OpenAI says Codex can be used to go from text to code for “essentially any programming task”, although they acknowledge that it can’t do all of these tasks very well. It may take rewording and tweaking of the instructions to get the Codex to do exactly what you want, and just like typing code yourself, sometimes an unexpected bug or two can creep in.

Not just for beginners: Being able to build games, apps, and websites in code without knowing it already or researching it all is obviously a big help for beginners who haven’t done it all before.

But moving from simple conversational text to code would also be a huge time saver for professional coders. During the WIRED demo, OpenAI CEO Sam Altman mentioned to Levy that Codex can do in seconds what would have taken him half an hour.

“It takes people who are already programmers and removes the heavy work,” Brockman told The Verge.

While Levy fears potential job loss and a cruel irony in people learning to code just as computers learn to code on their own, Code.org founder Hadi Partovi thinks automating parts boring to improve the teaching of computers, making it possible to train dreamers.

Being able to go from text to code can not only make life easier for programmers, but also open the door to people who don’t know the crucial language of the field: software.

“As coding rote work becomes easier, computer science education can focus on higher-level computational thinking concepts such as designing interfaces, algorithms, and data structures,” Partovi told Levy. (It’s like how modern art schools work; they don’t teach students how to paint, in itself.)

Text-to-code could be seen as the next step in streamlining computer programming, Zaremba told The Verge. What started as esoteric physical punch cards evolved into Short Code, BASIC, C, Python, Javascript and all of the hundreds of digital programming languages.

“These programming languages, they started to look like English, using vocabulary like ‘print’ or ‘exit’ and so more people became able to program,” he said.

“Each of these stages represents increasingly advanced programming languages. And we believe Codex brings computers closer to humans, allowing them to speak English rather than machine code.

Test, adjust and cultivate: Codex is still in private beta testing since its release in August. This period will be crucial in training Codex to be as competent as possible; the current iteration can only pass around 37% of tasks, Levy reported.

“OpenAI Codex empowers computers to better understand people’s intent, which can empower everyone to do more with computers.”

Open AI

Eventually, Codex can not only help programmers, but act as an interface — it can already run Windows, Spotify, and Google Calendar, The Verge reported.

“OpenAI Codex empowers computers to better understand people’s intent,” OpenAI wrote, “which can empower everyone to do more with computers.”

We would love to hear from you! If you have a comment about this article or have a tip for a future Freethink story, please email us at [email protected]

Comments are closed.