Codex AI Enters Private Beta
In the NLP-As-A-Service explosion OpenAI just keeps winning headlines. Headlines in an algorithmic world where headlines are fixed to some extent. So what about talking code? The new Codex API in private beta apparently allows people to translate words into code.
However we know AI coding for lay people is still likely years if not decades away, so what’s with the buzz? OpenAI keeps ringing the self-glamorizing alarm bells as if it were DeepMind, so what’s fact and fiction here? All this headline hype is probably nothing, for the Microsoft backed NLP firm.
If anything OpenAI is showing us some of the limits of NLP models. Codex is a descendent of GPT-3, a massive deep learning language model released last year, and it’s really not clear if this will ever be useful. What has Microsoft bought for $1 billion exactly with OpenAI?
The complexity of deep learning models is often measured by the number of parameters, still the idea of coding with an AI partner is sort of nonsensical. Will AI be able to auto-complete on the fly and make software engineers more productive? What is even the end game here?
Codex is best thought of as OpenAI’s versatile language engine, GPT-3, but trained only on code instead of ordinary written material. But there’s not much indication this is commercially viable or even remotely helpful to software programmers.
Many of these AI firms start off as not for profit and morph into crazy promises. OpenAI literally is promising companies the stars. When they say about Codex: Proficient in more than a dozen programming languages, Codex can now interpret simple commands in natural language and execute them on the user’s behalf—making it possible to build a natural language interface to existing applications. We are now inviting businesses and developers to build on top of OpenAI Codex through our API.
So you want me to buy access to an API to help you train it to be remotely useful? These “early products” aren’t even real products at all, yet. Codex is more of a next-step product for OpenAI, rather than something completely new. It builds on Copilot, a tool for use with Microsoft’s GitHub code repository.
It still doesn’t answer how OpenAI will become profitable or a useful firm for Microsoft to eventually acquire. It seems these NLP firms are easily able to replicate the work of each other, from China, Germany to Israel.
Is Codex Even a Viable Product with Real Market Fit?
Can APIs like Codex help software development actually to move forward? Sorry for being a bit skeptical. “Programming is about having a vision and dividing it into chunks, then actually making code for those pieces,” Brockman explained to TechCrunch. The intention with Codex is to let coders spend more time on the first part than the second.
So by that definition, Codex will help software developers become grunts in the software development process. The history of OpenAI does not inspire confidence in fidelity to a vision of AI for Good, not at all. OpenAI simply started out as yet another nonprofit entity in 2015 and changed to what it described as a “capped profit” entity in 2019—a move the company claimed would help it get more funding from investors. BigTech backing, dubious claims of products, what’s next?
OpenAI will be lucky even to be able to keep up with its competitors and China. Microsoft lending GitHub for OpenAI’s products is a bit like an unfair advantage. It’s pretty shady. On its announcement page, OpenAI says that it is releasing the API for Codex in a private beta to start and also notes that the company is inviting developers and businesses to give it a try.
Oh yes, let’s try your Microsoft project that makes software engineering feel even more like slavery. Slavery to AI, now that’s a good future for OpenMind. Just help invent the Overmind (an AI being) that helps humanity right its course while you are at it.
Better than copilot and an improved version of itself, Codex is graduating to Beta. All hail the NLP not for profit firms copying each other and getting acquired by BigTech.