[Tutor] louis renton

[Mats Wichmann]

On 1/22/23 17:50, Oscar Benjamin wrote:
> On Mon, 23 Jan 2023 at 00:35, Mats Wichmann <mats at wichmann.us> wrote:

> I think StackOverflow's reasons were in part very particular to their
> situation. From what I understand some people had promoted the idea
> that anyone could use ChatGPT to quickly gain "reputation". Then a lot
> of people blindly posted outputs from ChatGPT that were not helpful so
> SO banned it. I've seen something similar here:
> https://github.com/sympy/sympy/issues/24524#issuecomment-1383545797
> That usage of ChatGPT is bad but isolated (I haven't seen any other
> occurrence) so not massively problematic in the way that I think it
> was for SO.

I think the situation is a little related to what was claimed earlier in 
this thread (about "googling for stuff"), which is the only reason I 
brought it up:  the answers may well be right, even of high quality. 
It's even possible that that's the case in a large majority of cases. 
But if you don't have the background to make that determination, how can 
you trust it?  There have been lots of examples as people play around 
with the AI of the system being essentially "faked out" and getting 
something pretty dramatically wrong.