The J.A.R.V.I.S AutoReply Protocol Initiative (JARPI) - Powered by ChatGPT

[o_e_l_e_o]

I don't think adding more AI generated text to the endless amount of spam will achieve anything. No one actually reads such replies anyway; they just get in the way of actual discussion. If we want to solve the spam problem, then we should just actually enforce the rules regarding spam and signature campaigns and start handing out temporary bans like the rules say we will.

[joker_josue]

If we want to solve the spam problem, then we should just actually enforce the rules regarding spam and signature campaigns and start handing out temporary bans like the rules say we will.

But, shouldn't campaign managers be more rigorous in the posts they count?
I can even accept that it is not always easy to check one by one the posts that campaign participants make. But, I believe that these users who have a very low level of quality in their posts are easily detectable.

[NotATether]

Who decides if a topic is low-quality or repetitive? And who decides if the community is tired of regurgitating a topic?

Good question. Obviously ChatGPT and AI in general is totally unpredictable for this task, so I've considered making a sort of algorithm to calculate a spam score, like the kind that would be used at Google.

I think it would be better to make a bot that replies to shitposts with a warning letting the poster know they are shitposting. Something like:

Warning: your post appears to have been written by an NPC. This question has been asked on the forum approximately 43 times already, with the most recent merited reply being the following:

We can prove that the number of provably lost bitcoin is 2,828.654 BTC, and this number can be safely removed from both the total and the circulating supplies. Anything which is not provably lost should not be considered removed from the supply.

Consider performing a simple Google or forum search before posting a question on the forum next time. Thank you for your cooperation.

In light of other comments here, I think it would be better to send that in a PM. It would be a turbocharged version of Bitcointalk moderation staff in that case (and would admittedly be more "automation" than AI).

I don't think adding more AI generated text to the endless amount of spam will achieve anything. No one actually reads such replies anyway; they just get in the way of actual discussion. If we want to solve the spam problem, then we should just actually enforce the rules regarding spam and signature campaigns and start handing out temporary bans like the rules say we will.

Agreed.

The target audience that is supposed to be reading these kind of posts are not so much Legendaries like you and I as it is for new members who've just arrived here <3 months ago.

It's still generic AI crap. Right now I'm struggling with a bunch of my writers because they think that AI content sounds interesting, it doesn't. There's no emotion, just mathematically correct sentences with a lot of words to explain something that could be explained with five bulletpoints. Reminds me of answers that politicians usually give.

Being something of a writer myself, I also agree with you that AI-generates or otherwise low-quality content written by humans can be crappy (unless the text written is of an extreme technical standard that will go into a technical topic).


Perhaps a userscript could be developed that automatically hides shitposts from threads (or even entire topics).

[BenCodie]

I think you have the right idea of using A.I. to positively impact the forum - I do agree with the others that auto-generating replies is probably not the best way. I think you are right that the Bitcoin discussion board has many repetitive topics. Recently, I have started to think I was having deja vu while scrolling through it.

Something that would be able to grab relevant threads based on title and content, list them with a percentage of relevancy and post the list in the first reply for repetitive topics to allow the topic starter and following readers to easily visit threads that are very similar and read the responses there; would be a nice adjustment to this idea and a neat addition to that board.

[dkbit98]

Sorry but I don't like this idea of using AI in bitcointalk forum at all.
There is always going to be some human being in forum who is willing to reply to any question, however stupid that question may be, and starting with one AI-bot could result with more of them in future.
This bot can make forum members even more lazy and more inactive, that is just one of the possible outcome of experiments like this.
Only AI that I would consider slightly useful for forum is one that could detect if posts have high chance of being AI generated, but it would probably make many mistakes.

[PowerGlove]

I'm personally against the idea of synthetic replies. I'd like to (maybe) see AI applied to curating good posts or hiding bad ones, but certainly not to adding even more crap to the forum. -1 from me.

On the topic of fighting shitposts, what I would love to see is a button that lets humans (with >= 500 earned merits) mark a post as "redundant":



When 10 or more distinct users have clicked it, the post should be hidden from view, unless you've gone into your options and opted in:



Campaign managers would (hopefully) leave this turned off, and would therefore not end up paying for those posts.

The details can be voted on (e.g. maybe 10 distinct users is too few or the threshold should be configurable, maybe 500 earned merits is not enough or too much, maybe only DT members should be able to see the button, etc.)

[o_e_l_e_o]

On the topic of fighting shitposts, what I would love to see is a button that lets humans (with >= 500 earned merits) mark a post as "redundant":

I can think of plenty of users with >500 merit who would quite happily abuse this and brigade another user's posts to have them all hidden, especially in the case of trust disputes. Also, such a feature would undoubtedly be misunderstood and used for posts the user dislikes or disagrees with, not just posts which are redundant.

[Sandra_hakeem]

It's such a copacetic response to the whole repetition that goes on and on, mostly in the B/H board... I see alot of things that shouldn't be repeated twice Being over-emphasized a couple of times too.. sometimes, i don't even blame 'em OPs Cus yeah, they're right to ask questions on what doesn't seem right to them...
Infact,I believe there's a lot more to be done....I'm not just a code freek nor some stylish programmer like Gloves or Atheter, but I'm sure it'll come to a time where the forum will have to make some milli-timeframe pop ups, or redirects on FAQS, to Auto-synchronize...
AIs, as suggested would do the job too, but what if the forum ends up with getting the whole AI theorem espoused ?? How well will the young-age take advantage of them??? (... rhetorical anyways..)

Sandra 🧑‍🦰

[PowerGlove]

I can think of plenty of users with >500 merit who would quite happily abuse this and brigade another user's posts to have them all hidden, especially in the case of trust disputes. Also, such a feature would undoubtedly be misunderstood and used for posts the user dislikes or disagrees with, not just posts which are redundant.

Do you really think users would collude like that? I mean, a threshold (of distinct and qualified members) has to be reached before the post is considered "redundant". The merit requirement (and/or DT status, mentioned at the end) is there to mitigate abuse (with farmed accounts, etc.)

Similar logic applies to people misunderstanding how to use the feature; their individual clicks won't do anything unless (a fair few) other members have already flagged that post as well.

It's a bit silly to criticize (rather than change) constants that were chosen while spitballing. If 500 merits is not enough, then make it 750. If 10 unique members is a threshold that's too easy to game, then make it 15...

Feels like you just read the line you quoted, and then quickly scanned the rest.

[joker_josue]

I can think of plenty of users with >500 merit who would quite happily abuse this and brigade another user's posts to have them all hidden, especially in the case of trust disputes. Also, such a feature would undoubtedly be misunderstood and used for posts the user dislikes or disagrees with, not just posts which are redundant.

Do you really think users would collude like that? I mean, a threshold (of distinct and qualified members) has to be reached before the post is considered "redundant". The merit requirement (and/or DT status, mentioned at the end) is there to mitigate abuse (with farmed accounts, etc.)

Other systems that the forum already has, such as trust, sometimes suffer abuse, let alone this one.

Then, the fact that the post is hidden in the topic, it will not disappear from the list of posts made by the user, which is probably the list that campaign managers analyze.

[NotATether]

On the topic of fighting shitposts, what I would love to see is a button that lets humans (with >= 500 earned merits) mark a post as "redundant":

I can think of plenty of users with >500 merit who would quite happily abuse this and brigade another user's posts to have them all hidden, especially in the case of trust disputes. Also, such a feature would undoubtedly be misunderstood and used for posts the user dislikes or disagrees with, not just posts which are redundant.

Now this is exactly the kind of thing you could use an AI for. They're neutral, don't get into trust disputes with anyone and never apply to join signature campaigns because they don't know what the hell that is.

Like I said, I'm starting to see AI as something that should be more augmented next to the post buttons than as replies themselves. The only problem is, now we are looking at full-blown infrastructure required to host such a bot (a userscript, and a server to go along with it, for something that basically is operated on a free basis).

[Poker Player]

You did not refer to anyone in particular, but as I am one of those who expressed my opinion, I feel called out, so I will answer you. Personally, I consider it an insult that you put us in the same category by citing the example of those people who hate Bitcoin from the bottom of their souls because it does not fit into their business philosophies.

I think I've been known to not hold my tongue if I want to insult someone in the forum, so your perception does not coincide with my intention.

Think of it this way: first, personal human interaction was replaced by electronic communication, and now we risk ending up only talking to electronics. Call me old-fashioned, but I don't like it. If I want to interact with a computer, I'll use a Search Engine or spambot myself.

I deny the premise, human interaction was not replaced by electronic communication. Or do you live in a cave and only interact with others by typing on your computer? Some human interaction was replaced by electronic communication in the areas where it was most productive and useful.

[LoyceV]

On the topic of fighting shitposts, what I would love to see is a button that lets humans (with >= 500 earned merits) mark a post as "redundant":

If it's redundant, why not report it to be deleted?

If 10 unique members is a threshold that's too easy to game, then make it 15...

If it takes 10 or more "hits" to hide a post, I'm not going to waste my time tagging them. It would be much better if it's linked to the Trust list, so based on the user's who's judgement you trust already.
Barely any user has more than 10 negative tags, it's very unlikely a single post will reach that much. Plus, it means at least 10 people have to actually read a redundant post, which is a massive waste of time.

Some human interaction was replaced by electronic communication in the areas where it was most productive and useful.

In my experience, human interaction is replaced by a cheaper system. Some companies now have a computerized phone menu when you call them, and many of the options lead to a dead end where you don't get to talk to a human. It's cheaper, but not helpful.

[joker_josue]

Now this is exactly the kind of thing you could use an AI for. They're neutral, don't get into trust disputes with anyone and never apply to join signature campaigns because they don't know what the hell that is.

But the fact that they are neutral, does not have the human capacity to understand everything that is involved in any post. So you can sometimes unfairly rate something negatively that it isn't, or rate something positively when you shouldn't.

It is true that the bot can improve these details, but humans have infinite probabilities, so something can look like one thing to an AI and be something totally different, which only human discernment can do.

I'm not against AI being able to help our day to day, but I think that it generates content on the forum, it distorts the purpose of the forum.


Another detail:
Here we are discussing a "good" bot that was intended to automatically answer repeated questions.
But, what would stop the emergence of "evil" bots that will create content simulating that it is human?

We have to be careful not to set precedents that then make everything more difficult to control.

[PowerGlove]

If it's redundant, why not report it to be deleted?

Yeah, I'm with you. But, people are lazy and being able to just quickly click one button (instead of having to click once, lose focus, enter some text, and then click again) will appeal to many more people, I think. Also, having precisely defined rules (e.g. if 9 of your peers agree(d) with you, then the post will be effectively "nuked" without further ceremony) is more satisfying (to me anyway) than being at the mercy of a hit-and-miss moderation policy.

If it takes 10 or more "hits" to hide a post, I'm not going to waste my time tagging them.

I had the same thought. I think it might work best if the post immediately (or maybe with a second, user-specified threshold) disappears from your own view, but only disappears from everyone else's when the global threshold has been reached. I could see myself clicking that button all the time.

It would be much better if it's linked to the Trust list, so based on the user's who's judgement you trust already.

Yep, or something similar, like a new "post curator" list.

Other systems that the forum already has, such as trust, sometimes suffer abuse, let alone this one.

That's true, but I'm not saying that this system won't be abused, I'm saying that it's possible to select the parameters in a way that would make that abuse impractical to execute. (Of course, there's a balancing act here: if you set the countermeasures too aggressively, then the feature won't work that well in practice, because too many people would have to click the button for the "threshold" to be crossed.)

Then, the fact that the post is hidden in the topic, it will not disappear from the list of posts made by the user, which is probably the list that campaign managers analyze.

When I said the post would be "hidden from view", I meant the user's post history, too. Although, I'm guessing some campaign managers might use tools that rely on post archives (like ninjastic.space), and that needs a slightly more complicated solution (maybe a new endpoint to get a list of "redundant" post IDs for a given user and time frame).

[o_e_l_e_o]

It's a bit silly to criticize (rather than change) constants that were chosen while spitballing. If 500 merits is not enough, then make it 750. If 10 unique members is a threshold that's too easy to game, then make it 15...

You've already picked up on my point here in your next post: Using numbers which are sufficiently high to make it resistant to abuse will also make it essentially useless.

I am also in general against any proposal which brings us closer to Reddit's "mob rule" upvote/downvote type system, where unpopular but factually accurate posts are often hidden from view.

[NotATether]

On the topic of fighting shitposts, what I would love to see is a button that lets humans (with >= 500 earned merits) mark a post as "redundant":

If it's redundant, why not report it to be deleted?

Because we can't rely on moderators deleting all low-quality posts.

We have ~20 staff versus 20 thousand shitposts per day.

Which is why such posts should be automatically classified and hidden from everyone who uses a particular userscript.

[PowerGlove]

@NotATether: I'm starting to see your point about letting an AI classify shitposts. But, I don't frequent super spammy boards, so most of my issue is actually sig spam on good boards. The best way to combat sig spam is to have campaign managers stop paying for it. A system like what I'm proposing could achieve that without them having to change anything about how they run their campaigns, the posts would just automatically stop being paid for (because the managers would never see them). In my estimation, "AI" can often just be replaced with the phrase "black box", and I don't like the idea of affecting people's earnings with something that isn't accountable, makes decisions unilaterally, and frequently gets things wrong.

I am also in general against any proposal which brings us closer to Reddit's "mob rule" upvote/downvote type system, where unpopular but factually accurate posts are often hidden from view.

Yep, very much agree! But, surely you can appreciate how what I've described so far has the makings of a workable system?

Are you really so confident that no combination of parameters will end up working?

Don't you think some kind of trusted "post curator" list (or something similar) is worth exploring? (Please don't take the shortcut of immediately asking: "Okay, but who decides who's a curator and who isn't?".)

Or, what if the community could see a log (either combined or separated) of each user's activity with regard to this system, so everything was done out in the open, and negative trust could be used as a tool to combat abusers?

[o_e_l_e_o]

Don't you think some kind of trusted "post curator" list (or something similar) is worth exploring?

Maybe if it was made up of a few selected users rather than just anyone with "greater than x merit" which is easily abused, but then that raises the question "Why not just appoint said users as moderators"?

and negative trust could be used as a tool to combat abusers?

This would not be a good use of the trust system.

[LoyceV]

If it takes 10 or more "hits" to hide a post, I'm not going to waste my time tagging them.

I had the same thought. I think it might work best if the post immediately (or maybe with a second, user-specified threshold) disappears from your own view

Why waste time clicking individual posts, if you can Ignore everything from that user with one click?
If you want to share responsibility of hiding posts, it may be better to bring back the glowing Ignore button. It was removed before I joined, and I guess it was abused, but it might work in combination with custom Trust lists.

If it's redundant, why not report it to be deleted?

Because we can't rely on moderators deleting all low-quality posts.

We have ~20 staff versus 20 thousand shitposts per day.

It's more like 5000 posts per day, and not all of them are shitposts. But, instead of automatically hiding them, why not hook up with MindlessElectron to remove them