ChatGPT?

[Lemming]

Could this be ChatGPT or similar?

spam post (the poster later edited another post into spam)

It looked suspicious, being a bit bland, but it's well formed and makes sensible reference to earlier posts.

It's not important but I was intrigued by the idea that spammers might now be using this trendy tech.

[CopenhagenCouple]

Oh yeah, that was written by some language model based tool or bit and that's not the only case. Don't have time right now to look for specific examples, but I think we've had bot posts like this on at least a monthly basis (although maybe not as good as this, but they're learning and getting better by the day).

Most I've suspected usually look like they're set up to ID and use a few key words or phrases, run a semantic search algorithm, plug info into language model and create a short post with one or very few paragraphs that look like a semi relevant reply to the previous post or posts in a thread.

One of the typical signs of this is that they never really engage, e. g. ask questions or reflect on the topics, but rather give some info / build on / extend the previous posts. If there are any "points of view", they are usually referred to along the lines of "many people...", "some sources...".

A very funny little curiosity I've also noticed, is that the user names of many of these post will be extremely similar to the one you've linked, consisting of two names written without space between, but with capitalized first letters, e.g. JimSmith, PeteSamson, JanSantiago (not actual user names except the last, which is from the one you've linked, just illustrative examples).

Maybe we should start a game, who's up for a round of "spot the bots"?

[Free Ben]

That’s very interesting, I read the post linked above the other day and couldn’t put my finger on what I found unusual about the phrasing.

Initial thought was it was someone who’s first language wasn’t English but it just didn’t feel that it was the case.

Anyone watch the series ‘Humans’ on channel 4 a few years back? Anytime I see something like this type of bot pointed out I think that’s where we’re headed!

[Not_the_monk]

Yeah agreed, I doubt its actually that specific AI but its definitely a bot. And yes I've also seen a few others over recent months, they seem to be relatively inoffensive compared to the sort forums always got plagued with in the past as long as the volume and intensity doesn't increase too much, but they are also gonna be a bit harder to find ways to filter without manual review if they become an actual problem you'd think.

Maybe we should start a game, who's up for a round of "spot the bots"?

Ooooh, I'll go first. Do I get extra points for 2 in 1, even if they're short? Last 2 there...

viewtopic.php?p=432914#p432914

[CopenhagenCouple]

Ooooh, I'll go first. Do I get extra points for 2 in 1, even if they're short? Last 2 there...

viewtopic.php?p=432914#p432914

We've got our first winner and deffo an extra point for the twofer, both of them with the spurious naming also!

[CopenhagenCouple]

Ooooh, I'll go first. Do I get extra points for 2 in 1, even if they're short? Last 2 there...

viewtopic.php?p=432914#p432914

We've got our first winner and deffo an extra point for the twofer, both of them with the spurious naming also!

Forgot to add, that those two also exhibit another clue that we've got a bot on our hands, bumping old topics, for no apparent reason and without adding anything to them...

[CopenhagenCouple]

Here's a couple that date back to 2020 and aren't that obvious, but I suspect they're also bots.

Last post in this thread: viewtopic.php?p=427900#p427900

And especially the last post in this thread (made by same user with "botty name", close to the other post: viewtopic.php?p=427898#p427898

The last one in particular has me suspecting something fishy is going on, looks straight up like a chat bit answer to a question on studies on the topic...

Any thoughts on this one?

[Free Ben]

My entry!

Last post in this thread:
viewtopic.php?t=37038

[Free Ben]

….
The last one in particular has me suspecting something fishy is going on, looks straight up like a chat bit answer to a question on studies on the topic...

Any thoughts on this one?

I’d agree with you on these
These users in the examples so far no longer exist which points to them being booted following a potential further post (since deleted) of straight up spam

[CopenhagenCouple]

My entry!

Last post in this thread:
viewtopic.php?t=37038

Bingo, agreed on that also... Same name pattern, although interestingly using "I have..." etc more than others. User also deleted later, probably as you've stated wrt my entry, because they later straight up spammed with some post that got flagged or caught by meister lems

[Lemming]

As you say, this particular form of spamming has been going on for some time (hopefully I delete most). What was special about this one was the sophistication of the first post.

The usual pattern is a short reply to an existing topic. After a few days the spammer returns and edits the post to insert the link they want to promote. This one followed the same pattern except that the first post was quite long and appeared to engage several aspects of earlier posts without actually repeating them.

Of the ones you suggest, "CilinisStromy" and "FawnWitmer" were both deactivated long ago. I can't remember why I didn't delete all the posts. I must've just deleted the edited one for some reason.

"MelinaTorres" and "AllanGomez" are interesting. Joined and posted over a month ago but not been back to edit, if that is indeed the intention.

[CopenhagenCouple]

As you say, this particular form of spamming has been going on for some time (hopefully I delete most). What was special about this one was the sophistication of the first post.

Agreed, the one you pointed out was a new level, they're definitely learning My guess is that the programmer has given some new instructions, such as "summarize and comment on..." (very simplified) , instead if a more simplified start instruction such as "find topic and post info from search".

Given that there are initial indicators of GPT4 possessing at least something that resembles General AI, I think we've only seen the very beginning of this problem. Heck, we might be having a very similar conversation with a bot, not knowing it, in the near future. Allan Turing might have something to say about that...

[CopenhagenCouple]

Picking up on previous comment that it probably wasn't ChatGPT specifically, that was used to generate the answer Lemming posted, I think I agree.

However, it was definitely it was definitely some large language model based bot very similar to or just an earlier version of the model used for the one's I posted regarding ptsd and CBD. Just for fun, I typed "ptsd and CBD therapeutic use" into the 3.5 version, here is about half of what it returned:

"There is growing interest in the potential therapeutic use of CBD for PTSD and other mental health conditions.

Some studies suggest that CBD may have anxiolytic (anti-anxiety) and antidepressant effects, which could potentially benefit individuals with PTSD. CBD may also help to improve sleep and reduce the severity of nightmares, which are common symptoms of PTSD.

However, more research is needed to fully understand the potential therapeutic benefits of CBD for PTSD. While some early studies have shown promising results, there are still many unanswered questions about the safety and efficacy of CBD for this condition. "

That is very close to the answer given in the one I linked, just more articulate. My coding skills are a bit too rusty to make an attempt, but I know scores of people who could write a script or something using chatgpt and posts from forums like this to produce exactly what we're seeing here.

Interesting times, did anyone say Roko's Basilisk?

P. S. ETA: all hail the ai overlord!

[Not_the_monk]

That's interesting stuff CC. And yes indeed Lemming, that first one was definitely more sophisticated than the other examples above.

This made me go look find one I remembered seeing before that was more like it, knew I'd seen it in a Barcelona post a while ago...

search.php?author_id=120522&sr=posts

The Barcelona one you can see it has picked up a bunch of keywords from the first post and repeated them. But it definitely is more text, like the original example, than they usually do.Their other post, the Franco one, is interesting for the couple of typos like "Marco" and how it expands on the topic, almost gives you pause to wonder but then also well, nobody actually talks like that lol.

[CopenhagenCouple]

Those two examples are a bit harder @Not_the_monk, but I think you are right on those as well. Odd that it would misspell names from previous post though, it's not like it has clumsy fingers...