Be aware of AI posts

[EmptyBottle]

My mind and GPTZero have flagged some posts as AI. And before one questions "what about false positives" (yes, they exist, as with false negatives)... the post I submitted to GPTZero had a few AI tells too:

Lack of I statements
Overly consistent paragraphs
Suspicious summarisation

There wasn't as much grandiose language as the other post I flagged, and I understand a user who wants to put their emotions into words can use AI, but I would like it disclosed upfront vs me having to use detection techniques (some I won't mention in this post), and GPTZero.

BE AWARE OF (undisclosed) AI POSTS everyone please, and do your own detection too.

PS: The post I talked about before was not found anymore, only fragments remain on my phone's jpg cache

 

[Pluto]

 

[Forever Sleep]

Maybe I'm being stupid here. Apologies. What do you think has happened/ What can actually happen? Can an AI make an account here to begin with and start posting? Or, does a human make an account and then, lets AI take over? Or, is it simply a human inputting stuff into AI and copy/ pasting the result?

What is the actual danger do you think? Were the posts borderline breaking the rules? Inciting suicide or something?

I do personally prefer it when members make it clear they are quoting from someone else or, asking AI. A former member used to copy a lot from other sources I believe or use AI, without making it clear it was other people's work. That tends to trouble me because I worked in an industry that really valued copyright and intellectual property. The posts themselves were really interesting though. Good subjects to discuss.

 

[EmptyBottle]

Maybe I'm being stupid here. Apologies. What do you think has happened/ What can actually happen? Can an AI make an account here to begin with and start posting? Or, does a human make an account and then, lets AI take over? Or, is it simply a human inputting stuff into AI and copy/ pasting the result?

What is the actual danger do you think? Were the posts borderline breaking the rules? Inciting suicide or something?

I do personally prefer it when members make it clear they are quoting from someone else or, asking AI. A former member used to copy a lot from other sources I believe or use AI, without making it clear it was other people's work. That tends to trouble me because I worked in an industry that really valued copyright and intellectual property. The posts themselves were really interesting though. Good subjects to discuss.

My concern (for now) is the humans making AI posts and not disclosing AI usage, would be troublesome if someone fails a CTB in a painful way, due to AI error.

The post luckily didn't break the rules, and yep, sometimes were interesting to read, but because AI hallucinates potentially dangerous stuff, and if the user posted to a method post, I could see harm occuring.

Someone else already made a dodgy post about gun CTB that would have lead to injury only if followed.

 

[Forever Sleep]

My concern (for now) is the humans making AI posts and not disclosing AI usage, would be troublesome if someone fails a CTB in a painful way, due to AI error.

The post luckily didn't break the rules, and yep, sometimes were interesting to read, but because AI hallucinates potentially dangerous stuff, and if the user posted to a method post, I could see harm occuring.

Someone else already made a dodgy post about gun CTB that would have lead to injury only if followed.

I definitely agree with you, that either reference to AI or other sources should be disclosed.

It always puzzles me as to why people would even rely heavily on asking random strangers for method advice. I always think members should try to find the source material themselves. It's here after all- they just need to take the time to read the megathreads. I guess some specifics aren't covered though.

Personally, I try to avoid giving method advice. Mostly because I simply don't know but I wonder if it could even be interpreted as 'assisting a suicide'. I definitely agree though- if AI method advice is sounding incorrect, that is troubling.

I guess ultimately though, any of us- AI or humans could be remembering things wrongly. I've seen the wrong weight units being used in discussing the SN method. So, I think all members should try to verify stuff before they try out something for real.

 

[EmptyBottle]

I definitely agree with you, that either reference to AI or other sources should be disclosed.

It always puzzles me as to why people would even rely heavily on asking random strangers for method advice. I always think members should try to find the source material themselves. It's here after all- they just need to take the time to read the megathreads. I guess some specifics aren't covered though.

Personally, I try to avoid giving method advice. Mostly because I simply don't know but I wonder if it could even be interpreted as 'assisting a suicide'. I definitely agree though- if AI method advice is sounding incorrect, that is troubling.

I guess ultimately though, any of us- AI or humans could be remembering things wrongly. I've seen the wrong weight units being used in discussing the SN method. So, I think all members should try to verify stuff before they try out something for real.

ohh yes humans make mistakes too... also, I think there should be a thread on google quoted search so ppl can search vs re-ask.

eg "sanctioned suicide" [search term]

since SaSu disabled new users ability to search onsite for some odd reason, we'll need to educate ppl about how to search the site indirectly... since it is indexed on Google, there is no point hiding the search bar, unless to make the Private forums unsearchable (but maybe SaSu search can exclude them anyway for newbies)

 

[Forever Sleep]

ohh yes humans make mistakes too... also, I think there should be a thread on google quoted search so ppl can search vs re-ask.

eg "sanctioned suicide" [search term]

since SaSu disabled new users ability to search onsite for some odd reason, we'll need to educate ppl about how to search the site indirectly... since it is indexed on Google, there is no point hiding the search bar, unless to make the Private forums unsearchable (but maybe SaSu search can exclude them anyway for newbies)

I'm not entirely sure why they hide the search bar for new users. But, I don't believe it takes that many posts to unlock it.

I'm just not sure really- on whether it is all that good to make it so easy for new users. I once made a post in the suggestions section that there should be more emphasis placed on the method resource compilation thread in the stickies- for new users.

Partly to try and stop the constant: 'How do I kill myself? What's the best method?' that we get pretty much weekly and which always results in the same answers- 'We can't tell you what to do. Look at the resources page and make the decision yourself' pretty much.

But then, another member pointed out- maybe one of the few 'safeguards' here is that it isn't that easy to begin with. Being forced to wait even a little bit perhaps gives us time to think. If we could create an account and have our method and sources within a matter of minutes- would that be good for everyone? Not sure really.

I'd tend to argue that anyone who goes through the process of creating an account here, waiting for it to be verified etc. doesn't really seem massively impulsive to me but, I don't know really. I think there's a fine line between a freedom of information and wrecklessly handing it out without truly knowing about the person. There's bound to still be minors on here- for one.

 

[hedezev4]

I use AI as a translator, because regular translators translate too badly and I have to correct a lot. But sometimes AI also changes the meaning of the translation, and I have to tell it what needs to be fixed. But I never ask it to write the text for me, I always write it myself in my native language, then send it to AI, read it in English, look for places where it changed the meaning, and correct. Also in work specifically with chatgpt it really annoys me how often it likes its long dashes (—). Sometimes it happened that only from the 3rd time, when I asked it to remove the long dashes, it did it.

And so I completely agree with you, AI very often makes mistakes especially in narrow specialized topics. Try to ask it anything about video games or an instruction how to do something in a program, it almost always will hallucinate.

So if someone used AI then it needs to be written. By the way I also saw that post about guns ctb, where there was no mention that it was AI (as far as I remember) and that was very wrong.

I think that the absence of search is quite ok, as already mentioned above, those who want to find will be able to bypass this limitation with the help of search engines, and those who did not really try to find information may delay their impulsive decision while waiting for this function to appear.

 

[EmptyBottle]

I use AI as a translator, because regular translators translate too badly and I have to correct a lot. But sometimes AI also changes the meaning of the translation, and I have to tell it what needs to be fixed. But I never ask it to write the text for me, I always write it myself in my native language, then send it to AI, read it in English, look for places where it changed the meaning, and correct. Also in work specifically with chatgpt it really annoys me how often it likes its long dashes (—). Sometimes it happened that only from the 3rd time, when I asked it to remove the long dashes, it did it.

And so I completely agree with you, AI very often makes mistakes especially in narrow specialized topics. Try to ask it anything about video games or an instruction how to do something in a program, it almost always will hallucinate.

So if someone used AI then it needs to be written. By the way I also saw that post about guns ctb, where there was no mention that it was AI (as far as I remember) and that was very wrong.

I think that the absence of search is quite ok, as already mentioned above, those who want to find will be able to bypass this limitation with the help of search engines, and those who did not really try to find information may delay their impulsive decision while waiting for this function to appear.

ohhh, good point on the impulsivity risks.

GPTZero is somehow able to tell the difference between polished with AI and pure AI. Tho some human signs like the use of I statements may remain, and the removal of the em dashes also suggests human curation.

I'm not entirely sure why they hide the search bar for new users. But, I don't believe it takes that many posts to unlock it.

I'm just not sure really- on whether it is all that good to make it so easy for new users. I once made a post in the suggestions section that there should be more emphasis placed on the method resource compilation thread in the stickies- for new users.

Partly to try and stop the constant: 'How do I kill myself? What's the best method?' that we get pretty much weekly and which always results in the same answers- 'We can't tell you what to do. Look at the resources page and make the decision yourself' pretty much.

But then, another member pointed out- maybe one of the few 'safeguards' here is that it isn't that easy to begin with. Being forced to wait even a little bit perhaps gives us time to think. If we could create an account and have our method and sources within a matter of minutes- would that be good for everyone? Not sure really.

I'd tend to argue that anyone who goes through the process of creating an account here, waiting for it to be verified etc. doesn't really seem massively impulsive to me but, I don't know really. I think there's a fine line between a freedom of information and wrecklessly handing it out without truly knowing about the person. There's bound to still be minors on here- for one.

ohh yes, adding some friction to CTB so it isn't

Contacting the server for information...
Retrieving method file 1...
Running method1.exe


and instead:

Contacting the server for information...
Querying the server...
Awaiting response...
Retrieving methods file 1 to 10...
Unable to automatically select out of [2,3,5,7]
Contacting the operator...
Deleted method files 1 to 10.

 

[Scythe]

Okay, let's say there's AIs here, for them to be here in the first place they'd need a human to register them, I highly doubt an AI can pass registration on its own.

What does the human registering the AI get from this? They can't post in large numbers cause they'd be suspicisous and the mods would notice, there's also already a ton of posts that can be viewed as a non-member if they just want to gather info on mentally ill ppl. So for this reason I highly doubt there's AI here, aside from unless some asshole wanted to spread misinformation. And judging from ur post it seems not all the posts flagged as AI were method related.

 

[EmptyBottle]

Okay, let's say there's AIs here, for them to be here in the first place they'd need a human to register them, I highly doubt an AI can pass registration on its own.

What does the human registering the AI get from this? They can't post in large numbers cause they'd be suspicisous and the mods would notice, there's also already a ton of posts that can be viewed as a non-member if they just want to gather info on mentally ill ppl. So for this reason I highly doubt there's AI here, aside from unless some asshole wanted to spread misinformation. And judging from ur post it seems not all the posts flagged as AI were method related.

true, at least for 2025, I hope it doesn't get to the point where AI infiltrates by itself... tho my concern is about humans posing AI as themselves (maybe not intentionally, if they don't know it is good etiquette to tag AI as AI)... and that causing issues like the gun post with AI (misinformation)... accidentally inserting pro-lifer arguments because they didn't read what the AI wrote and embarassing themselves / annoying others unintentionally, etc.