Re:Goodbye to Google Code Jam
by schrodinger on Friday March 24, 2023 @10:13PM
(#35298469)
Sometimes things that have been discussed repeatedly on this forum and are a basic consensus don't need a citation, and if you'd like one you may do your own research. This is a lazy and unhelpful response.
[ Parent
| on HN ]
I haven’t seen a clear answer on this actually - there was some confusion if the training data contained solutions for older problems, and if it underperforms on previously unseen ones. Plus it still seemed to underperform on medium and hard problems? Does anyone in the know have a summary on this?
Re:Goodbye to Google Code Jam
by camdenreslink on Friday March 24, 2023 @10:19PM
(#35298510)
I think it has only been shown that ChatGPT can solve easy/some medium leetcode style questions. I doubt it can solve difficult problems that would give human competitive programmers a hard time (for now anyway...)
Re:Goodbye to Google Code Jam
by nozzlegear on Friday March 24, 2023 @10:18PM
(#35298507)
Sometimes things that have been discussed as possibilities on this forum begin to be repeated and taken as fact. One might say that this forum has hallucinated the capabilities of this bot beyond its actual capacity, and the hallucination grows daily. A citation is indeed needed.
Re:Goodbye to Google Code Jam
by unethical_ban on Friday March 24, 2023 @10:38PM
(#35298665)
GPT4 has been out for what, weeks? We aren't talking established theorems here.
Oh, I looked and didn't find anything. I thought about writing something about how I thought it wasn't true but decided it would be better to just ask if they had a reference for it. also, it's an extraordinary claim, so pretending like it's something everybody already knows is pretty weird.