Hi Andreas,
I completely agree with the point you make about correctness. I personally feel 
LLMs have a use for exploration and bug finding. Even if I agree completely 
with the motivations of the GCD. 

>Just yesterday I asked several LLMs a simple question about a project
>I know:
>"What is the default precision of MPFR?". Several of them replied that
>there is none, claiming that it must be given by the user on variable
>initialisation (both wrong); one went on to state that there is a default
>rounding mode (which is wrong, it needs to be given by the user at each
>function invocation). And it gave code snippets that initialised variables,
>but did not free them.

I think this example mischaracterizes a bit these tools, I suggest you to use 
them in agent mode (they work fine inside a guix shell container, I usually 
share only /tmp or a subset of it). I am not familiar with MPFR, I assume it's 
a git repository. I would probably have asked: "clone the source of MPFR, study 
it and tell me the default precision of MPFR. Provide evidence for any 
statement you make about the code in the file_name:line_number format."

This has worked well for me and gave a correct answer with evidence. Of course 
one has always to assume that they are hallucinating, but asking for evidence 
makes checking their statements easier.

cheers,
giacomo

Reply via email to