If some LLM somewhere managed to produce a coherent and _correct_ explanation of how it worked inside, but no human ever saw it, would it matter?
We know how to apply LLms to problems, which is a subtly different thing to understanding how they do what they do. It's similar to fire: I can cook using fire, but I don't really understand how fire _works_. Heat+oxygen+fuel, sure, but what goes on chemically? I dunno. Doesn't stop me using it. (Pretty sure _humanity_ knows how fire works, though).
"AI recombines successfully" is a phrase that makes me think of the human corpus as a genome and a particular AI response as a child of the parent prompt, with the thinking a model does in response to the prompt basically being a form of information sex.
Machine code is code like any other code. Sure, often not easy to read or modify, but not impossible. I'd start with a disassembler, a debugger, possibly a decompiler and lots of patience. Given appropriate time and pay I might even consider it a fun job.
I suppose it depends on the code you're trying to understand. Depending on its size, style (therefore, ability for a human or for a decompiler to understand it) and other factors, the best blend of tools can change.
reply