AI guidelines

[mindrust]

The real question is, what happens when the AI starts producing text outputs that are indistinguishable from a real person's post? That's already possible. If the person knows some basic English he can read the AI response and re-write the same thing using his own words but there will also be other, better "humanizer" tools out there. (they are already available as far as I know).

[nutildah]

The real question is, what happens when the AI starts producing text outputs that are indistinguishable from a real person's post? That's already possible. If the person knows some basic English he can read the AI response and re-write the same thing using his own words but there will also be other, better "humanizer" tools out there. (they are already available as far as I know).

My previous take on that was if they're going through enough effort to have to re-write the AI post, taking almost as much time as it would to just write a post (or maybe more), just to avoid the detectors then I guess thats OK. But the recent output of AI shitposts has taken a turn for the bizarre... and they're not any easier to read. artificial shitposting is the new thing.

[joker_josue]

The real question is, what happens when the AI starts producing text outputs that are indistinguishable from a real person's post? That's already possible. If the person knows some basic English he can read the AI response and re-write the same thing using his own words but there will also be other, better "humanizer" tools out there. (they are already available as far as I know).

My previous take on that was if they're going through enough effort to have to re-write the AI post, taking almost as much time as it would to just write a post (or maybe more), just to avoid the detectors then I guess thats OK. But the recent output of AI shitposts has taken a turn for the bizarre... and they're not any easier to read. artificial shitposting is the new thing.

If the user rewrites the AI text, the text ceases to be from AI, becoming a text from the user himself.

The problem is with those who copy-paste.
But gradually AI systems will produce texts that are more difficult to detect, and it all depends on the requests made by the person.

Unfortunately, it will be difficult to understand if a user is writing with AI or not, in a single post.
It will be necessary to analyze more than one post. We will have more work, but only this way.

[nutildah]

Going beyond the issue of AI and addressing the nature of ban-able posting habits:

It might be a good idea to come up with another metric to measure the worthiness of a post. Something that captures fringe AI cases (which we will see more and more of), meaningless posts in general, and obvious spam. I was thinking something along the lines of a "good faith" indicator:

Was the post written in a genuine attempt to address a relevant topic of discussion? (it doesn't have to be a good attempt, just genuine)? If not, how many of the user's last 40 (or so) posts are also like this?

How can we tell if its not?

• Its written by AI or someone else
• It fails to address the topic of conversation
• Its contents aren't substantially different from replies before it, or the author's other posts.
• Its apparent the author hasn't read any of the replies
• Its obvious trolling or spam

There's probably other telltale signs of posts not written in good faith but this is a start.

[Vod]

There's probably other telltale signs of posts not written in good faith but this is a start.

How do you feel about marking/quoting all AI responses?

Example:  What is life to you?

If the user rewrites the AI text, the text ceases to be from AI, becoming a text from the user himself.

I usually verify/edit text before I post it as a quote.

[nutildah]

How do you feel about marking/quoting all AI responses?

Example:  What is life to you?

Personally I think its fine, so long as its not in every post. That would probably get annoying, if someone had to consult their chatbot for every post they write.

In your particular post, the source of the text is labeled, and the output is relevant to the discussion at hand, at least somewhat.

[Vod]

at least somewhat.

That's why I edit the text - to avoid discussion creep.  I removed most of the garbage stats to just show women carry far more responsibility in rearing children.   But I don't want to hijack this thread for my example. so I'll just stop typing.

[MarryWithBTC]

The real question is, what happens when the AI starts producing text outputs that are indistinguishable from a real person's post? That's already possible. If the person knows some basic English he can read the AI response and re-write the same thing using his own words but there will also be other, better "humanizer" tools out there. (they are already available as far as I know).

I think this is very well captured by theymos in the last paragraph of the AI guidelines. 
You talked about reading AI responses and re-writing it using their own words. This should be a secondary concern, the primary concern should be how the AI got its prompt, what kind of prompt was given it. Was it a prompt to get an idea? if yes, it's fair but if it is a prompt of cluelessness asking AI to write a whole text, this will be disastrous.

However, theymos recommends against ever relying on AI writing advice very much.

You can use AI to suggest grammar/style improvements, or to give you ideas, with you manually implementing those suggestions; but avoid directly copying more than a few words from the AI. However, I recommend against ever relying on AI writing advice very much, since mixed in with its good suggestions it'll usually also give bad suggestions which will just erase all personality from your writing. Better your writing be flawed and authentically-you than "correct" but stripped of any soul.