Policy on use of ChatGPT • page 1/6 • Lichess Feedback • lichess.org

We've seen a number of postings in the forum recently which are clearly produced by ChatGPT. Are other people concerned about this development?

My own feelings are a bit mixed. I understand that ChatGPT is an incredibly useful tool to formulate text, and it allows some people to take part when they would otherwise be unable to do so. Writing what you want to say does not come easy for everyone, especially if it's in a language which is not your first. However, is the output, which is typically lengthy and in a style which is unnatural in a discussion, appropriate for the forum? To me it doesn't feel authentic. And I like to communicate with people, not with AI.

What about blog postings? Is the use of ChatGPT okay when formulating them? I'm a bit more liberal on that personally, because blogs aren't discussions (though the blogger may allow discussions about them on the forum).

I don't know of any policy on this platform regarding the use of AI tools to formulate text. Perhaps we need one?

We've seen a number of postings in the forum recently which are clearly produced by ChatGPT. Are other people concerned about this development? My own feelings are a bit mixed. I understand that ChatGPT is an incredibly useful tool to formulate text, and it allows some people to take part when they would otherwise be unable to do so. Writing what you want to say does not come easy for everyone, especially if it's in a language which is not your first. However, is the output, which is typically lengthy and in a style which is unnatural in a discussion, appropriate for the forum? To me it doesn't feel authentic. And I like to communicate with people, not with AI. What about blog postings? Is the use of ChatGPT okay when formulating them? I'm a bit more liberal on that personally, because blogs aren't discussions (though the blogger may allow discussions about them on the forum). I don't know of any policy on this platform regarding the use of AI tools to formulate text. Perhaps we need one?

TotalNoob69

#2

Some people cannot articulate their ideas in English as well as a native speaker. ChatGPT does a pretty good job translating stuff, even from bad English to good English. Besides, what are you afraid of? That a machine is going to generate a good forum post? Because bad forum posts we've learned to live with already...

glbert

edited

#3

@TotalNoob69 said in #2:

Besides, what are you afraid of? That a machine is going to generate a good forum post?

chatgpt generates dog shit forum posts though, not good ones. they are very boring to read because of their unnatural and verbose style. as i mentioned here: https://lichess.org/forum/lichess-feedback/lichess-anti-cheating-feature#6 i would much rather read bad english by a non-native (quite like my own posts) than the slippery marketing nonsense produced by chatgpt. and in fact i skip anything that reads like chatgpt anyway.

so what i am afraid of is a world in which all the overuse of chatbots leads to forums that are unreadable to me.

@TotalNoob69 said in #2: > Besides, what are you afraid of? That a machine is going to generate a good forum post? chatgpt generates dog shit forum posts though, not good ones. they are very boring to read because of their unnatural and verbose style. as i mentioned here: https://lichess.org/forum/lichess-feedback/lichess-anti-cheating-feature#6 i would much rather read bad english by a non-native (quite like my own posts) than the slippery marketing nonsense produced by chatgpt. and in fact i skip anything that reads like chatgpt anyway. so what i am afraid of is a world in which all the overuse of chatbots leads to forums that are unreadable to me.

TotalNoob69

#4

I, for one, welcome our AI overlords. Law is for humans, context is for AIs.

corvusmellori

#5

Trying to police if forum posts are written with the help of generative AI seems like a losing battle and not a great use of very limited resources. The post speaks for itself. If it's an inappropriate post, regardless of its origin, it should be moderated; otherwise not.

glbert

#6

@TotalNoob69 said in #4:

Law is for humans, context is for AIs.

did you use an ai to come up with that? makes no sense whatsoever to me.

@TotalNoob69 said in #4: > Law is for humans, context is for AIs. did you use an ai to come up with that? makes no sense whatsoever to me.

glbert

#7

@corvusmellori said in #5:

Trying to police if forum posts are written with the help of generative AI seems like a losing battle and not a great use of very limited resources.

it does not necessarily have to be policed. we can use social stigma just fine. if we refuse to read ai created posts, fewer people will post them.

@corvusmellori said in #5: > Trying to police if forum posts are written with the help of generative AI seems like a losing battle and not a great use of very limited resources. it does not necessarily have to be policed. we can use social stigma just fine. if we refuse to read ai created posts, fewer people will post them.

Cedur216

#8

At some earlier point it was said that chatgpt's overlong walls of text count as spam?

petri999

#9

recognizing AI post is kinda hard. If feel it too verbose when generating just tell the thing to be less verbose and it obeys. A friend mine selling consulting on GTP applications showed an example for text the AI was supposed to tell if text AI generated. And the it got them other way round i.e human written was judged to be AI generate.

So maybe not so to tell always.

recognizing AI post is kinda hard. If feel it too verbose when generating just tell the thing to be less verbose and it obeys. A friend mine selling consulting on GTP applications showed an example for text the AI was supposed to tell if text AI generated. And the it got them other way round i.e human written was judged to be AI generate. So maybe not so to tell always.

Brian-E

#10

I certainly agree with those who are making the point that detecting what is AI and what is genuinely written by the poster is extremely difficult, perhaps a bit similar in some ways to detecting engine cheating in games, and would require huge effort for relatively little gain.

But as glbert mentions in #7 a policy of discouraging AI generated postings doesn't have to be rigorously enforced by moderators. Not all of the items on the forum etiquette page https://lichess.org/page/forum-etiquette result in immediate moderator action when they are infringed.

To respond to @TotalNoob69 who asks in #2:

what are you afraid of?
I'm afraid of a situation where ChatGPT-generated postings are dominating the topics. I think those postings are generally conversation-stoppers. Those who are now already using ChatGPT regularly may be joined by others who discover the ease with which they can generate walls of text with no effort, and that could bring in the demise of what is, at present, a good, lively discussion space.

I certainly agree with those who are making the point that detecting what is AI and what is genuinely written by the poster is extremely difficult, perhaps a bit similar in some ways to detecting engine cheating in games, and would require huge effort for relatively little gain. But as glbert mentions in #7 a policy of discouraging AI generated postings doesn't have to be rigorously enforced by moderators. Not all of the items on the forum etiquette page https://lichess.org/page/forum-etiquette result in immediate moderator action when they are infringed. To respond to @TotalNoob69 who asks in #2: > what are you afraid of? I'm afraid of a situation where ChatGPT-generated postings are dominating the topics. I think those postings are generally conversation-stoppers. Those who are now already using ChatGPT regularly may be joined by others who discover the ease with which they can generate walls of text with no effort, and that could bring in the demise of what is, at present, a good, lively discussion space.