Topic: The J.A.R.V.I.S AutoReply Protocol Initiative (JARPI) - Powered by ChatGPT - page 2. (Read 902 times)

For the JARPI bot, no matter how hard you spin those texts manually, they won't be original, and have you considered the amount of posts this bot will be posting per day since it triggers every time someone asks a question, or are there limitations I'm not aware of, this AI thing is getting out of hand, these tools would render some of our education sector useless, resulting in the loss of jobs if nothing is done now.



You don't expect the admin to ban someone for 6 nasty posts, do you? Like you said AI written post is difficult to prove if it is plagiarism or not because it is nearly impossible to cite the main source, so they are automatically classified as spam and get deleted overtime, there are tons of users with over 100 deleted spam posts who are still here going about their unusual spamming business. Ignoring him is the best you can do in this scenario; we shouldn't make the forum tough and uncomfortable for newbies; I'm sure most of us didn't start out well, and everyone learns with time.

Remember when you asked the Admin to ban DdmrDdmr for spamming when he was a newbie? Today, he is regarded as the best thing that has ever happened to the forum. patience!

I have no problem if Moderators want to give someone a second chance. Even if they think someone deserves several chances, that's okay. But, it should vary from member to member. You have to see who the member is. If it's someone prominent member or someone who has the potential to contribute positively. You may give them several chances to prove themselves. I saw ETFBitcoin reported several plagiarisms of the same user and he is still doing the same shit every day.

A permanent ban is too harsh with the ban evading rule. But, when you are giving them several chances and you see he continues doing the same shit. You should think about it. If you see the same user doing it again and again, you shouldn't forgive them every time. Users like Jamesborn, Godlovesyou, and many more deserves punishment for what they doing.

Like Edcrypto91. I reported 4 of his posts, and 6 of them got deleted. But he didn't get banned, which is disappointing. Maybe he got a temporary ban, but I can't know that. Chat AI spam is plagiarism, but without clear source it's hard to prove. If shitting all over the tech boards isn't enough to get this user banned, reporting him feels like a waste of time. He's now on my ignore list, which makes one less user to report his posts.

I can't blame you for not knowing the solution. I understand that theymos don't want to put more restrictions and keep the forum as free as possible. But Forum Mods can update the forum rules over time, especially for AI. Some people use AI to write posts; when you ask them more about it, they cannot tell you in detail. They don't even know what they are talking about. There are a bunch of tools that they can utilize so they don't get caught. At first, they write content using GPT, then they use paraphrase tools so you cannot detect AI Writing using another AI.


That's why bounty participants paid for (For spam).

Personally, I do not see any objection to the use of artificial intelligence robots modified manually, this will not eliminate the role of members in providing their own touch on the topics or responses that they write.

This is similar to when someone uses Google Translator and then manually revises the translation so that it appears human rather than machine.

But on the other hand, locking the topic when using JARPI will make most of the discussions brief and deprive many members of responses.

But is there a better way to do this, other than manually?
The best way to assess the quality of a post is by a human. Especially because the concept of quality can vary from person to person, and even if there are guidelines, only a human being will be able to evaluate it in the best way.

This is really not an easy thing, but we have to debate ideas to find ideas/solutions that allow us to reduce this level of spam/poor quality.

Hmm... maybe "reserved" posters could get warnings or be kicked out of their campaign as a punishment?


Makes me wonder whether there's such a thing as an "academic score" for measuring the substance of texts. The idea being, the same university-level standards that are applied against plagiarism currently are used to score a post based on several criteria like originality and informativeness.

And people who don't even manage to pass the score get no bonus (and ChatGPT, lacking any emotion at all, would score a zero and get ejected).

For example, in my campaign there is a bonus that goes to the best posters. Although it is quite a manual process, maybe there is some heuristic some researcher has come up with for this in the last 100 years.

You would undoubtedly have 5+ posts saying "Reserved" show up immediately on every single topic posted in the technical boards.

You have to be really careful when thinking up incentives, because they can easily backfire and make the problem they were intended to solve worse rather than better.

It's like that story about trying to get the rat population under control during the French occupation of Hanoi. The authorities incentivized citizens to kill rats (by offering them a small reward for each tail they brought in), and although it did work for a short time, citizens then started leaving the rats alive (after taking their tails), hoping that they would breed and produce more tails for them to collect the bounty on. There was even rat farming going on, to take better advantage of the rewards. Authorities realized what was happening and stopped paying for the tails, but the damage was done and they now had an even bigger rat problem than when they started...

If you add a financial incentive to getting your post in early, then I think that would ultimately decrease the quality of the first few replies. There's already a kind of "traffic jam" at the beginning of some topics (usually questions) with the first few replies trying to compete for merit. I often don't bother to post in topics like that because it would take me too long to do the question any real justice, and I know that by the time I'm finished with my reply there'll already be a bunch of (pretty lazy) "fast" answers that (taken as a whole) cover most of what I was going to say. It would actually be pretty cool if there was a way to call "dibs" on a topic and stop other people from posting "rushed" answers while you work on a comprehensive one.

That's true, but I am speaking of larger incentives: an additional $50/week for making an early constructive answer is nothing to pass over.

Using AI to automatically flag / report spam / nonsense threads and posts does seem like a useful application of this technology. It would need to be a somewhat purpose-built AI though, trained on actual Bitcointalk data. Not as easy as 'throw it into ChatGPT', sadly, but probably much more effective.

I don't understand why my opinion gets ignored but if you don't believe me, just check and analyze the situation yourself. So, I'll tell you why actually repetitive questions are asked on this forum:
There are a lot of people who post just for the sake of signature campaign earnings. There are threads in every boards that require some knowledge and mental work to answer. For someone, who doesn't speak English and didn't have access on proper education and comes from a very poor country, it is getting difficult to post in these threads and it will take a lot of time, so here is what they do: One newbie registers on Bitcointalk and creates threads that have been answered thousand times but doesn't show up on first page of board and at the same time it's easy to answer, for example threads like this: Is money the root of all evil?, Bitcoin vs Stock market, Will bitcoin replace banks?, Will bitcoin end world poverty?.
These threads are easy to abuse, you just always side with Bitcoin, even with nonsense text and no one complains. So, the scheme is following: One user creates thread that is easy to spam and hundreds of users post in this thread. Then this one user creates another thread again and again. This one user is not alone, there are other users who follow his path and finally tons of spammy threads are created that feeds spammy users.

Just check it yourself, this is something that needs to be fixed. I am not against what you do but to be honest, an AI can do nothing in this case. One positive thing it can bring is to report some nonsense posts or threads to moderators.

I can definitely understand it. Since it writes so well and human-like, it is easy to be fooled by it and believe that the text contents are correct, as well. Also, most people have no prior experience with AI and thus it is fascinating to interact with one for the first time.
I'm sure when researchers interacted with the first somewhat working AIs over 20 years ago for the very first time, they must have felt a similar feeling to the normal user logging into ChatGPT today. AI has always had this 'almost there' kind of feeling in my opinion, where it feels like it's about to be an omnipotent, or at least omniscient thing that's going to revolutionize everything.

Getting the job in a campaign already is the incentive to make useful posts

It is a multi-layered problem.

If only all first replies were constructed as carefully as that, then there would hardly be a visible spam problem, as any userscript can just filter all spam from Page 2 or reply #10 onwards or something.


Well since we are on bitcointalk.org, it is possible to make some incentive for campaign members to post a well constructed and detailed (non-AI generated) as the first/second/third reply of a question thread, and that will reduce the amount of spam people have to wade through to find an answer ultimately.

But I'm not a campaign manager, I am a "hacker", and thus I must work with what I have.

Uh.....yeah.  This is as far as I've gotten in reading this thread, but I'm going to be consistent and reply before I finish.

Not only would I not want to see replies automatically generated in response to idiotic or just plain common questions, but I also don't think it's going to do anything to reduce the number of sig/bounty shitposters chiming in with equally idiotic replies.  If you have a thread started with a question like "what's the best hardware wallet?" and and the first response from a member is the most comprehensive comparison of all the hardware wallets out there with a detailed description of each one, there's no question in my mind that the following replies will be half garbage and half commenting on whatever.  AI isn't going to solve any of this forum's problems.

WORD UP.

I've been watching some videos about this stuff lately, and sweet satan is it scary as hell.  I think we're all fucked in about five years.

I found it funny that you mentioned me, because I had zero experience with these new fangled AI systems.  But, this discussion got me curious so I just spent an hour playing with ChatGPT to get an idea of how it would respond.  I agree that it would be fairly easy for me to distinguish, or at least I would also like to believe so.  The the example I posted below, for instance; it's accurate and coherent, for sure, but contains no personality.  The precision is too mechanical, almost predictable in it's encyclopedic accuracy.  Soulless, for lack of a better term.

ChatGPT just repeats stuff like parrot and it has no sense of good and bad, I recently watches one guy asking him questions and ChatGPT had to restrict him access in the end because he was making conflicting statements all the time.
My point is that this AI is not available all the time, it can't work without human controller, and I have no idea why people are fascinated with it so much.
We the people can survive and function just fine without ChatGPT, not the other way around.

I appreciate the effort and do like the idea in the sense that questions actually repeat themselves quite often. When I encounter such topics, I try to quickly find a similar topic (similar to StackOverflow topic suggestions) and refer the user to it.
I think making a bot do something like this (or at least one that assists users in this manner, e.g. as a browser extension that uses TryNinja's search and suggests similar topics) would be more useful than having a language model write full replies.

My main issue with ChatGPT is that (as mentioned before) is a language model. Its entire purpose is writing text that looks / reads good. Not writing correct text. Therefore I'd avoid using ChatGPT for anything but creating text (or maybe code) templates to fill out yourself. Like a letter template.

I don't know the solution At least there's much less spam than we had years ago, thanks to the Merit system.
And there's this:
(click to read the rest of the post)

I'm in the same boat. But in that case, just giving those users Mod-powers would solve it for everyone.

Spam wouldn't exist if it wouldn't be profitable.