This shows a grave misunderstanding of what compilers and LLMs are. They're fundamentally opposite concepts.
Compilers are about optimizing abstract code down to the most efficient representation possible for some hardware. LLMs are about wasting petaflops (made possible by compiler engineers) to produce random statements that don't have any static guarantees.
How can you trust that the compiler has written the most efficient assembly, if you’re not double checking it by hand?
Jokes aside, I understand your point.
In the history of computing, LLMs and compilers are closer than one might think.
Compilers weren’t first created to optimize “abstract code down to the most efficient” assembly as possible, even if that is the goal of a compiler writer today.
Compilers were created to enable the use of higher-level languages. Abstraction, efficiency, portability, error reduction, and most importantly: saving time.
They allowed humans to create more software, faster.
This shows a grave misunderstanding of what compilers and LLMs are. They're fundamentally opposite concepts.
Compilers are about optimizing abstract code down to the most efficient representation possible for some hardware. LLMs are about wasting petaflops (made possible by compiler engineers) to produce random statements that don't have any static guarantees.