AI Spam Report Reference Thread

[Satofan44]

I don't really know what you are saying dude, I'm really surprised you said my post are generated by ai but guess what, you can check the results I got below 👇

You should try understand that most of this ai detectors don't give accurate results because they also make mistakes.

Guess what, in case you don't know, you can check this thread below

You literally did not even read or understand my post when you responded with this shitpost. I have proven that you use AI, there is nothing you can say. No detector will help you anymore. You deserve a sig. ban like 99% of the users in signature campaigns even if this case is not confirmed to be AI, nutildah is not working fast enough. @nutildah work faster, add an extra shift for the first few weeks. Get to it.

[TokenTikas]

User: Mr_Brilliant$

Good catch but I want to mention one more thing. Most of his writing gets detected by chatboard detector. Even after reporting Mr_Brilliant$ and bringing AI-detected posts to attention, he is still not stopping writing with AI. It is suspicious that a user like this suddenly started using AI. For that reason, they should be kept on the target list and their activities should be monitored.

Thoughtful one.. Going into a business because there is actual demand for it, rather than because other people are doing it, is the better approach to business in my opinion too..

Something else I will add is that knowledge is also very important.. Before venturing into any business, it is important to know how it works, the risks involved, the market, and what you are getting yourself into.. because If you go into any business without proper knowledge, there is a high chance that you will end up making costly mistakes that could have been avoided..

Undetectable AI: 99%
GPTZero: 100%
StealthWriter: 100%

Running is still consistently posting AI-generated content despite being called out for it.

[nutildah]

No detectors were used or needed for this detection, I was working with the AI during the time when I was reading some responses in this thread and this order of words caught my attention. Then I checked and replicated the output, which is better than a generator -- I find them increasingly unreliable as many have started to sell "humanize" services, because of which they have an incentive to falsely label text as AI. Merit history also confirms this is a farmed account. @nutildah signature ban please.

It would set a terrible precedent to start sigbanning accounts solely based on suspicion without corroborating evidence.

You appear to be right, but its not worth accusing accounts here unless you're willing to put them through the process outlined in the OP. That goes for everybody. The process exists for a reason. If you're not satisfied with the current detectors, find a better one, or make your own.

I don't really know what you are saying dude, I'm really surprised you said my post are generated by ai but guess what, you can check the results I got below 👇

Your posts are almost certainly "AI inspired", at the very least. I recommend you and all other accounts currently doing what you're doing to stop now. If you think you will get away with it forever, you are sorely mistaken.

You are taking AI output and running it through a shittifier. (There's no such thing as a "humanizer" because AI is incapable of outputting human text.)

Here's an example. Original post:

For me, I will say that not every action that happens during or after war means that everything is being planned from the start. You see when the tension reduces even for a little period of time, businesses and countries do usually uses the opportunity protect their own interests. During this time some will strengthens the alliance, some will trade while others will prepare should in case the conflict starts again. Meanwhile it do not automatically mean or prove that whole conflict thing is orchestrated, but it shows that every country is trying to secure the best outcome for itself in a very tense situation.

Sapling: 0.4% Fake

After removing unnecessary filler words, correcting punctuation and spelling, the results are vastly different:

Not every action that happens during or after war means that everything is being planned from the start. When tension reduces even for a little period of time, businesses and countries usually use the opportunity protect their own interests. During this time some will strengthen the alliance, some will trade, while others will prepare should in case the conflict starts again. Meanwhile, it does not automatically mean or prove that whole conflict is orchestrated, but it shows that every country is trying to secure the best outcome for itself in a very tense situation.

Sapling: 99.9% Fake

However, this is not enough on its own to issue any kind of action based on AI use. I am sure there are better detectors out there that recognize shittified AI text right away. We should not confidently label this account as an AI poster until it can be more clearly demonstrated.

[Satofan44]

It would set a terrible precedent to start sigbanning accounts solely based on suspicion without corroborating evidence.

You appear to be right, but its not worth accusing accounts here unless you're willing to put them through the process outlined in the OP. That goes for everybody. The process exists for a reason. If you're not satisfied with the current detectors, find a better one, or make your own.

It would were it not now validated through 2 other methods by both you and me -- how is that not "corroborating evidence"? There is no objective reason that would make the detection tools given here or method the only ways to "corroborate evidence". By explaining how I did it it I provided a means for others to reproduce it. You can try to find a better detector if you want, because from what I can see the detectors that are used here have started to recognize even writing that predates the internet as AI. Sometimes it works, sometimes it doesn't and would fail to detect the post of the user as not AI even when it is clearly AI. That is why I delivered the example user with another method. Query the thread where the reply was written to see what AI would say, and then compare it or subtract things like you did -- both work.

However, this is not enough on its own to issue any kind of action based on AI use. I am sure there are better detectors out there that recognize shittified AI text right away. We should not confidently label this account as an AI poster until it can be more clearly demonstrated.

We can skip the AI labeling since it deserves a sig. ban anyway. You can use suspicions cast on users due to other behaviors (such as this one) to put priority on the account for evaluation for a sig. ban -- you are not supposed to act on anything without investigating it. I should have been clearer initially. The detectors are mostly a waste of time when an user deserves a signature ban, the step can be completely skipped if they check the other boxes.

You deserve a sig. ban like 99% of the users in signature campaigns even if this case is not confirmed to be AI, nutildah is not working fast enough. @nutildah work faster, add an extra shift for the first few weeks. Get to it.

The user Mhizlove has completely generic, AI-padded, pattern-confirmed posts. It does not need an AI tag, it needs a sig. ban. It is just one of the many accounts that are farmed up. Often you can find that they have earned all of their merits in the recent 120 days to rank up by tracking signature campaign applications (while the accounts are still fresh, obviously as time passes this changes), even if they are complete shitposters. Meanwhile you can find other users that have posts that are significantly better but they barely earn any merit. That is the pattern, it is so obvious that the level of enforcement here is a joke.

https://bitcointalk.org/index.php?action=profile;u=3726250;sa=showPosts

There's not even 1 post that is not a shitpost.

[nutildah]

Trust me, I will get around to at least some of that in the future. Right now the backlog only includes the most poorly-rounded accounts, which overlaps with some AI users, but not all.

I will likely end up building my own AI detector to better capture the different types of AI posters on the forum:

- template AI (low level)
- shittified AI (medium level)
- Google Translate AI (high level, sometimes hard to tell if thoughts expressed are original)

It is half done already, but not specific enough as it currently stands.

This is the value of the detectors:

A full member who posts using AI.
BattleDog

Shamir backup is basically splitting your seed into pieces, then require a minimum number of pieces to rebuild it. So if you pick 5 shares with a threshold of 3, you can lose up to 2 shares and still recover, but a thief would need to steal 3 of them before they can do anything. It's not multisig and it doesn't change your wallet on-chain, it's just a fancier way of storing the same secret.

People use it when they want redundancy across locations or some light "steal one paper and you get nothing" protection.

The tradeoff is complexity: most losses I've seen over the years came from people getting cute with backups and then not being able to reconstruct them under stress. If you do try it, do a full recovery test once (on a fresh device, with no cameras around) and consider something boring like 2-of-3 with each share stored separately...... otherwise a well-protected standard seed (and maybe a passphrase if you understand it) is honestly fine for most folks.

Cheers

StealthWriter: 100% AI detected
undetectable: 60% AI detected
GPTZero AI: 100% AI detected

This guy fooled absolutely everybody, except for the detectors.

He probably could have had a nice career as an intelligent AI reply guy if he hadn't joined a sig campaign early on. Then his real motivations became much clearer.

Speaking of that, my system was trained (in part) on data from accounts with pre-existing sig bans, which includes BattleDog and a few other AI posters who had already been sigbanned at the time.