> Before the advent of these models, the main argument against automating these tasks was that machines can’t think creatively.
Really? For me it's always been, and hasn't changed since LLMs have been released, that it's much harder to explain in enough details what you want to an AI and get the correct result than to actually code the correct result yourself.
Prompting is like a programming language but keywords have a random outcome, it's 100% inferior to simply writing the code itself.
LLMs will help with tooling around writing the code, like static analysis does, but it'll never ever replace writing code.
Really? For me it's always been, and hasn't changed since LLMs have been released, that it's much harder to explain in enough details what you want to an AI and get the correct result than to actually code the correct result yourself.
Prompting is like a programming language but keywords have a random outcome, it's 100% inferior to simply writing the code itself.
LLMs will help with tooling around writing the code, like static analysis does, but it'll never ever replace writing code.