My question got an answer that used ChatGPT but still got 22 upvotes. Is this usage allowed on Puzzling Stack Exchange?
$\begingroup$
$\endgroup$
5
-
1$\begingroup$ As someone who's been playing around with ChatGPT a bit, this definitely should be addressed early, and should be expanded to AI in general. We're going to see rapid improvements in AI very soon and it'll have a big impact on everything $\endgroup$– Beastly GerbilCommented Dec 20, 2022 at 0:36
-
7$\begingroup$ Hi. Please don't keep updating the number of votes that answer got so far. It adds nothing valuable to the post, but unnecessarily bumps the question. $\endgroup$– ACBCommented Dec 25, 2022 at 9:52
-
$\begingroup$ Thanks. I would stop. $\endgroup$– mathlanderCommented Dec 25, 2022 at 15:30
-
2$\begingroup$ So, all the answers to your question used an online tool to solve it, and you do not like one of them because it has "ai" in the URL? $\endgroup$– wimiCommented Jan 11, 2023 at 7:36
-
$\begingroup$ Really, please stop editing the question to change the number of votes. You could just do ">20" if the out-of-datedness bothers you. $\endgroup$– bobbleCommented Mar 22 at 14:56
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
4
Why ChatGPT and AI answers should be banned... mostly
It is my opinion that pasting an answer from an AI model should not be allowed in general as it goes against the spirit of puzzling.
(Note this is more general, I think the use of ChatGPT in the linked question is fine and would fall under 'exceptions' below)
Why AI answers go against the nature of PSE
When we boil it down, the true nature of this site is for users to create puzzles for others to solve. While the quickest answer is rewarded with a checkmark, the purpose of setting the puzzle is not to get the quickest answer, but to challenge another human being with your creation.
Using AI to solve a puzzle, and then just copying and pasting the answer, takes away all the human aspect of puzzling, and defies the purpose of setting the puzzle in the first place. Puzzle setters aren't trying to create something that is difficult for an AI to solve, but for another human.
Exceptions
It may sometime be appropriate to use AI to either help with, or indeed answer, a question. Not every question on PSE is a typical puzzle, and there are other question types where it would be completely appropriate to use AI.
Such examples might be the linked question, where the user is just looking for a list of answers to a problem, or where a user is stuck on a puzzle, such as a Sudoku or wooden block game, and can't find the next step, in which case it is already common practice to research how to help.
It may also be appropriate to use it for research, similar to how a normal search engine would be used, to help deduce a part of the puzzle. From my experience of ChatGPT, it already can give better and more useful information to certain questions than a normal google search.
How can we know?
At the end of the day, it won't be possible to know if AI has been used to solve the puzzle, so any ban won't be able to be enforced.
Stack Overflow have already banned ChatGPT answers for coding questions, more on the basis that it can generate wrong information, and the issue of enforcement has been brought up there too.
This is an issue that already affects PSE however, so there will be little difference. There is no way to know whether a user has 'cheated' to solve your puzzle, by using online tools or code to answer the question.
Does it matter too much?
Right now, AI will struggle to solve the majority of puzzles on the site anyway.
It is possible to potentially solve short letter and number games, but good luck asking it to approach the door...
AI will obviously see some major improvements, and its capabilities will increase drastically - but I still don't see it being able to solve a lot of the current puzzles on the site anyway.
There is also the fact that the vast majority of users wouldn't want to use AI to solve a puzzle anyway. They are here for fun, for a challenge, and at the end of the day cheating with AI for a couple of extra internet points is not really worth it.
A summary
AI answers should be banned, or at least heavily discouraged, unless it is an appropriate scenario. This won't be enforceable, but for the most part it won't need to be - few users will want to use it to cheat and it (at least currently) won't be capable of solving the majority of puzzles on the site.
-
2$\begingroup$ The "challenging another human" aspect may not be all that relevant to the puzzle that's linked in the question. The question explicitly allows computer answers and the currently accepted answer is just typing a regex into a computer tool and looking at what output it gives. $\endgroup$– JafeCommented Dec 20, 2022 at 1:15
-
$\begingroup$ @Jafe oh I completely agree for the linked question, should probably clarify that in the answer - this answers more general. The linked question IMO would fall under the exception category, as the question is just looking for a list of answers $\endgroup$ Commented Dec 20, 2022 at 2:13
-
2$\begingroup$ SO banned ChatGPT on the basis that it can generate a wrong answer, and Puzzling SE banned ChatGPT on the basis that it can generate a correct answer. $\endgroup$ Commented Dec 26, 2022 at 3:33
-
4$\begingroup$ @FranckDernoncourt yep pretty much. You have to bear in mind that PSE and SO are very very different sites - SO doesn't want wrong answers which is understandable and PSE just doesn't want people using computers to solve the puzzles for them. I think both are fair $\endgroup$ Commented Dec 26, 2022 at 18:44
$\begingroup$
$\endgroup$
Well I think if you couldn't use AI then using Google search is out. Also no auto correct for spelling or grammar. That doesn't seem reasonable.
What is the purpose? The fear is that it spoils it for other people. Surely that is already against the spirit if what we are doing here and there are other rules against it.
ChatGPT might challenge us to come up with more original puzzles as essentially all it does is mash together stuff it's already read.
