2 ms·
I don't care that much for the commentary about who will or will not make it. After all, I think that if AGI is really just around the bend, then we don't have
by jchw 2y ago
I don't care that much for the commentary about who will or will not make it. After all, I think that if AGI is really just around the bend, then we don't have to worry about programming jobs: they probably won't exist. But if it's not, then these tools will probably continue to improve at a rate that makes what we have today irrelevant. I believe that will happen with all of the image generation knowledge people are accumulating with Stable Diffusion derived models.
That said, forgetting all of the AI doom or hype, this seems frankly incredibly cool. I haven't managed to find a great use case for AI code generation where it hasn't failed miserably (except for minor boilerplate uses, like with IntelliJ's LLM auto complete for Java), but actually, generating FFI libraries seems like a pretty kickass use case for automation. I dig this. I will definitely try it.
- XCSme 2y ago> we don't have to worry about programming jobs Maybe not writing the code itself, but creating a product/business still requires deciding how things will work, what technologies to use, what's the public API, etc. So instead of writing the code to do it, you still need someone to ask the AI to do it: "create a to-do list with login system, that has e2e encrpytion and stores the data in a NoSQL database", etc. So you still need some programming knowledge to use it, otherwise you don't know what to ask for.
- chii 2y ago> otherwise you don't know what to ask for there's a perfectly working LLM for which you can query the answer! Ask it to tell you what are the important aspects of product XYZ.