Imagine typing “make kitty fall from the sky” and watching code generate that makes your cat image fall from the top of your screen. It’s raining cats instantly!
OpenAI has unveiled a new Codex, an “AI system” that translates your plain English text into code.
Last month we shared the launch of Copilot, the AI-enabled pair programmer that completes your code. Think of how Gmail suggests the end of your sentences in emails. Like that, but with code.
Copilot was created by Github and OpenAI using an earlier version of Codex, but the makers had only given us a peek at their Codex research at the time. Now they’ve released an updated version and an API, making it available for free in private beta.
Codex, a descendant of GPT-3, is trained on code — notably, virtually all the public code on Github along with other repositories. Codex has the natural language understanding GPT-3 does, which means when you write “fall from the sky” not only does it know what you mean, it knows to accelerate speed to emulate an item falling from the sky.
Codex is most capable in Python but also proficient in over a dozen languages from JavaScript to Ruby. Not-a-shocker: It’s not perfect. Techies and journalists are quick to test its limits, digging deeper into stacked, complex commands. Another criticism worth noting is that some coders take issue with the fact that Codex is trained on their code repositories.
OpenAI seems overall unconcerned and bullish about how this technology will help coders. The launch blog explains:
“Once a programmer knows what to build, the act of writing code can be thought of as (1) breaking a problem down into simpler problems, and (2) mapping those simple problems to existing code (libraries, APIs, or functions) that already exist. The latter activity is probably the least fun part of programming (and the highest barrier to entry), and it’s where OpenAI Codex excels most.”
We have a lot of GPT-3 fans in the house, so we’re not surprised that so far, the Product Hunt community seems to agree.
“Everybody will truly have an opportunity to be a maker,” Slava Bobrov wrote.
What do you think?
- “Last year I was doing web & UI-based freelance work for brands like Nike. I learned a lot but wasn't making money when I wasn't working. Now, I dabble with digital products I can build once and sell repeatedly. Ask me anything.”
- OnlyFans for investors? Quooore 2.0 launched its updated social investing app.
- If what you need is buried in nested folders, check out the launch of Playbook.
Companies can have treasure troves of data, but that data is worthless if they don’t know how to analyze it. That’s why Mixpanel is partnering with Product School, the leader in Product Management training, to bring you a free online Products Analytics Certification course so anyone can master the “ins and outs” of product analytics. This self-paced course will teach you how to build behavioral metrics from scratch and unpack complex topics like acquisition metrics, retention metrics, and product experimentation.
