New generation of (chatgpt(?) bots

[aka13]

I am really curious seeing new bots emerge.
I have seen similar stuff on phbb forums for years, but they were dumber, and injected spam links on their first post.

For a couple of months though I see new generation bots posting on-topic comments in different unrelated threads.

I hope you have game testers that replace the public crowd from initial development that continue to tell you if you're doing "bad or stupid". Not only professional game testers (they often don't really care about a game but only about their given task on paper) but also ordinary players who play the game just for entertainment, because for us players, the entertainment value is the most important property of any game.

Absolutely! It is critical to combine inputs from expert game testers with everyday players. Professional testers provide controlled evaluations, whilst regular players provide honest input on the entertainment value of the game. Balancing these views ensures that a game not only achieves technical standards but also resonates with its audience, resulting in a satisfying experience.

Hello,

adding as the only thing "Landfill" on "Tiles" in a Deconstruction Planner, leaving the list mode to "Whitelist" and changing the drop box to "Only", while nothing is set in "Entities", won't let me remove any landfill tiles.

Is there any other way to remove accidentally placed landfill tiles?

You can manually deconstruct landfill tiles by selecting them and giving the deconstruction order. Simply hover over a landfill tile and press the deconstruction key (usually the right mouse button) or use the deconstruction planner to mark the landfill tiles for removal.

They still are fairly obviously non-human, too real to be real, and the choice of topics is not as logical as a human would have chosen, but it's funny to see the latest generational tech weaponized for spamming purposes.
I wonder what the new anti-spam has to be, to detect these, they are almost undetectable even by humans. I also wonder what the endgame is, upping the post count until they can spam links?

Removed hidden spam from the quoted messages. boskid

[Qon]

I wonder what the new anti-spam has to be, to detect these, they are almost undetectable even by humans. I also wonder what the endgame is, upping the post count until they can spam links?

Removed hidden spam from the quoted messages. boskid

Seems like they are spamming links, just not visibly? Since boskid removed links.
If the color is set same as background color, you should still see it when quoting though...

[SoShootMe]

I wonder what the new anti-spam has to be, to detect these, they are almost undetectable even by humans. I also wonder what the endgame is, upping the post count until they can spam links?

There was one I reported recently which (from boskid's comment about hidden links) was similar to those you quoted - the only purpose I can see is to publish the link as a particularly scummy SEO technique. It was deleted so I am going from memory but it asked variations of questions that had already been answered in the thread. The two examples you gave are I think a little more obvious, but perhaps that's just because I'm reading it with that view: something about the form of the first seems "typical ChatGPT" to me and the second is superficially reasonable but actually wrong so... I guess the same .

Generating something vaguely approximating human is still computationally (and ultimately financially) expensive; orders of magnitude more than posting the resulting text. This could be the saving grace as I think it will stay relatively expensive, but it may not need much (or anything) to be sufficiently more effective that it is nonetheless worthwhile, in which case it could easily become a menace.

The best defence is to limit/prevent it being generated in the first place, which is in the interests of organisations providing such services. Or maybe not, as the second best defence is probably to use the same technology to detect it. A bit like an antivirus software vendor creating viruses.

[Koub]

Sorry for the spambots I may have let through, I will try to step up my bot detection game.
Maybe it's because I'm not a native English speaker, I have to admit they are far from obvious to me so far.

[aka13]

I wonder what the new anti-spam has to be, to detect these, they are almost undetectable even by humans. I also wonder what the endgame is, upping the post count until they can spam links?

Removed hidden spam from the quoted messages. boskid

Seems like they are spamming links, just not visibly? Since boskid removed links.
If the color is set same as background color, you should still see it when quoting though...

Holy shit they outsmarted me . If they never wanted me to click on them, than it must be indeed SEO. I never bothered looking exactly what I am quoting, just copy-pasted.

Sorry for the spambots I may have let through, I will try to step up my bot detection game.
Maybe it's because I'm not a native English speaker, I have to admit they are far from obvious to me so far.

I already said so multiple times, but I gladly will repeat myself, you are doing a superior job in filtering/policing the forum for years, don't worry.
I just never noticed a link being hidden in those messages, so I never reported those before. Now the quote-tool will be a handy weapon to detect those, now that I understand what exactly they are doing.
I am also nowhere near to a native speaker, I just get the "uncanny valley" effect from generational bots (I wonder how long it will last though )

[SoShootMe]

Sorry for the spambots I may have let through, I will try to step up my bot detection game.
Maybe it's because I'm not a native English speaker, I have to admit they are far from obvious to me so far.

I am a native English speaker: the examples in this thread are not obvious to me either and there are definitely no complaints here. I think most, including most native English speakers, would say the same.

There is something (that I am unsure how to describe) I can see in the style, but it is less difficult to see it when the idea is already in one's mind. Like identifying the artist of a previously unseen painting or the composer of some previously unheard music, I think this needs experience combined with "guidance" from contextual clues and can nonetheless lead to the wrong conclusion.

The example I mentioned was a little different to the ones here in that it asked questions related to the topic. It revived an old thread, and was a first post - both good clues in general but a first post asking questions, apparently having diligently searched the forum and found a related topic but missing that they were already (but not quite directly) answered, didn't immediately raise alarm bells. Only when the link became obvious in quoted text was the game up.

In short: keep up the good work .

[Qon]

Sorry for the spambots I may have let through, I will try to step up my bot detection game.
Maybe it's because I'm not a native English speaker, I have to admit they are far from obvious to me so far.

I don't think your english is a problem. Just get accustomed how ChatGPT speaks and it will become easier.

For the first example, that sentence is very typical of ChatGPT, agreeing, saying you should seek expert opinion, always adding in opposing viewpoint to not be too extreme, not actually saying anything on the topic. Waffling like a coward in a political speech or doing damage control PR stuff.

Pretty much a GPT copy pasta with some context words swapped in to make it sound a bit like it is related to the topic.

The second isn't that, but it has no understanding of the game so it just makes something that sounds like what could fit. It just ignored the question being about landfill tiles which are permanent. It ignored the settings in the planner in the quetsion, just answering like ... a regular forum user ... (now maybe we can ban all users for botting if they don't add anything to the topic xD)