Bitcoin Forum

An interesting read.
Please read the whole article before going on through my post.

OpenAI's GPT-3 may be the biggest thing since bitcoin (https://maraoz.com/2020/07/18/openai-gpt3/)


Well, we may have finally shed light on the AI nature of some users.
/j
Actually, I think the implication here might be important to consider.

Think only about the implication on the level of spam two bots can generate discussing together, or the problem for campaign managers, of even for paid services like bumping services of marketing solutions.
Two boys discussing and shilling a new project can push social engineering to a whole new level.
This kind of software might be too sophisticated to be intercepted or proven to be used with an high enough degree of certainty.

I think it is something needs to be addressed.

Definitely scary to think about. Although, I’m sure the bots have been here with us for a long time, just not as sophisticated as OpenAI’s new GPT-3. ;D

For example, take a look at this guy's posting history: Subbir (https://bitcointalk.org/index.php?action=profile;u=2691534;sa=showPosts)

I have read the article and to be honest I did not notice for a single time that the article could be written by an AI. If such posts can be generated from AI, I think we have some in our forum. The article was correctly written and is far above grade in term of quality from most of our forum users.
Some days ago, I think someone had reported user who were posting randomly. May be they are the creation of GPT-3.

Maybe the AIs will evolve to the point where they generate a more stimulating conversation than traditional spammers.  Could you imagine how much it must sting if you're an actual person trying to earn merit and a damn script comes along that manages to earn more because it makes more sense than you do.    :D

That might cause big trouble in our community (or even other forums) because we aren't able to determine the characteristic of posting on this AI bot since the bot can able to generate text that relevant to the topic. Sound like prediction answer.

But I noticed that there are some awkward words that the AI bot has been used. And the creator was right, probably we can determine in this way.
https://i.imgur.com/7aXNB0f.png

If someone reply or making a post something like this conscious to unconscious probably they're bot or spammer.

Some say I've been doing that for years.

With the fast developments in IT and machine learning, bots will be upgrade to smarter AI-bots and I believe that they will be able to wisely enough to create posts and join discussions. It is the thing and change to come to our forum, obviously. You can see the same thing with Google translator. Years ago, translation products from Google translator are terrible in quality but now its quality is acceptable enough and can be used for people can not translate original documents by themselves.

Do you think we can avoid such AI-bots evasion here? I don't think so.

But to say AI-bots can damage or destroy our forum, I don't believe they can.

We have:

• Merit system (https://bitcointalk.org/index.php?topic=2818350.0): to prevent spammers, and bots (even with AI-bots) to rank up. If needed, merit system's upgrades can do better to prevent them.
• Bump score (https://bitcointalk.org/index.php?topic=5183553.0): to prevent bad effects from bump services. Even if tehy are AI-bots, they can not beat the bump score and the algorithm of bump score can be upgraded too.
• Combined effects from merit system and bump score: they can not rank up, their bump score are nearly zero and (if >0) can not increase.
• Reports from community can help: to delete spam posts, bump posts, and nuke related accounts.
• Avatar and signature disabled can help: I think AI-bots are smart enough to realize those changes and stop posting

Basically, who create bots, even AI-bots? Human, so human can beat bots.

---

I don't like Facebook but they have a feature to disable rights to share a livestream if finding something suspicious on exponential shares in short period of time. Maybe sometimes findings are false ones but it is helpful.

I did a subtle reference to the quasi-AI nature of your parent account in my OP, actually: a suspect many have arisen in the past.

I lose brain cells everytime I read one of his posts lul.

Well, the article is of a great quality compared to the badly written and badly translated spam we have here in great quantities.
Which, on a sadly sarcastic note means that AI posting here instead of those guys would increase the quality of posts.
Unfortunately this also means that it'll be harder to find out which poster is genuine human and which is an AI posting high quality, well written... made up trash.

Unless theymos gonna make or use AI which can detect whether a message is written by AI/not, there's nothing much we could do aside from reporting posts that doesn't make sense.

There's simpler & good horrible solution for this problem, which forcing user to solve CAPTCHA every time he made/edit message.

I assume this a brand new release?

There's a Reddit sub where only GPT bots are allowed to post. That is fucking awful. The sentences are complete and coherent but the whole place echoes with absolute and overwhelming emptiness.

I'm impressed that the linked article was entirely AI. There were a few things that jarred but overall it's a (much) higher standard than most posts here. I can't really see how this can be resisted or identified, especially if it continues to be refined.

The years of google translate or generally empty shitposting has kind of softened us up to be more receptive to this too. I presume there is some gaping hole where this system can be confused or defeated. We'll have to wait for it to be uncovered.



That would be the end of my relationship with this place. But there'll be a bot for that too no doubt.

I have no doubt AI research is going to change things, but bitcointalk forum may also change in time (I hope).
One thing we could see is some way to prove we are humans with some kind of sybil tests, or with new way to log in the forum.
Lot of bitcoins are spent for new forum, and I think forum proof-of-person system can also be considered for future.

Using bots isn't even against the rules.

Me too, and it makes me very curious where it gets its input data. Basically, it just made up some lies! What's the point in that?

I do see potential though, we'll have thousands of different AIs compete to join signature campaigns managed by other AIs. Or, thinking bigger, an AI can write those boring work reports. Then another AI can read it, and in the end nobody has to do anything anymore. Especially with the recent work from home, your own cloud AI can apply for thousands of jobs per minute pretending to be you, and since you don't have to show up in person anyway, your AI can maintain dozens of jobs :D

AIs are much better at it than humans. Try to solve Google reCAPTCHA using Tor within 5 minutes. It's painstaking for humans, while an AI has all the time it needs.

I for one welcome our new AI overlords. Finally I can ask "is CSW the real Satoshi" and get an honest answer. Or most likely get an irrelevant salad of Google search results and tag the fucking bot with a neutral... actually even red trust would be probably ok since I don't think that trading with a bot would be advisable at all but I digress.

I haven't seen any AI capable of having an actual opinion, let alone style or personality, and the article hasn't convinced me. But it might be able to fool a signature campaign long enough to get paid so there is definitely a niche here.

It's just getting warmed up. We'll have to wait and see. The opinion bit is the one they'll be looking to refine no doubt.

This was the all AI Reddit sub I referred to - https://www.reddit.com/r/SubSimulatorGPT2/

It's a truly painful read but strangely reminiscent of an awful lot of posts on this very forum.

If the article posted by OP is really written by using AI, then I can only say "Houston, we have a problem". I believe that the predecessors of GPT-3 are in use and have found practical application, especially on forums and some social networks, and that we will soon get literally virtual characters in the virtual world which will very faithfully imitate real persons.

If I have well understood the whole procedure for AI to write an article, it comes down to the setup of basic parameters on the basis of which then GPT-3 writes that article. I looked at some of Manuel Araoz's tweets and here's what it looks like in practice.

https://i.imgur.com/0ZVRK2V.jpg
https://twitter.com/maraoz/status/1284581042175836163

The pricing (https://inferkit.com/docs/pricing) is still a lot less than even low-paying (Bitcoin) signature campaigns.
In the long run, the cost will probably go down with cheaper computing power (or the quality goes up).

We already know what the answer is - https://bitcointalk.org/index.php?topic=5124947.0

https://i.imgur.com/s7WqDGP.jpg?1

Or perhaps everyone will have to read their posts out loud on video, but keep an eye out for deep fake Tom Cruises if they forget to reset it.

Once again another reason for signature vacation time.

Every week  flip a coin and either shut signatures off or turn them back on.

Makes bots and ai have less reason to post if there are no signature revenues.


Unfortunately our forum is existing for the wrong reasons.  

Signature revenue is a shitty reason to exist.
Spam is worse.
Scammers are the worst.
Trolling fit in there somewhere but not as bad as a scammer.


At loyce how about scraping a signature campaign thread and estimate what it pays each week.

I often complain about signatures and have asked for signature vacation time.
So here is another reason to try it out.

I don't really pay attention to them but how many mindless and mediocre posters are left in campaigns at this point? It's not like days of yore where anything could sign up. You have to be pretty convincing to get in one.

Almost all of the crappiest posts are sig free. I presume they'll use something like this in the hope of ranking up to get in one. Not sure how far they'll get.

It could be mildly successful because the are no negative merits. So if they manage to find a way to earn even 1 merit per 100 near-zero-effort AI posts they would eventually rank up to full member or whatever is the minimum these days for a signature campaign.

The weak spot in this, like with word spinners, is that just a single suspicion can unravel the whole thing. Campaign managers, fellow campaign participants, merit senders etc will eventually notice that something is not quite right with the 10k-post user who streams generic texts without showing any lasting engagement with other users.

OTOH if the AI can beat that then more power to it... them... will it have a preference for certain pronouns I wonder.

Unlike plagarisers there'll never be a smoking gun which makes things a bit harder to execute. I guess you can build up a collective amount of disengagement but the owner of the account will be incentivised to drop in every now and then and act like a reactive community member.

I vote for us to close our eyes and wish for it to go away. The alternative might be a bit hard to take.

I can understand that you are against signature campaigns, but I cannot agree that the situation is so bad that we have to ban them. When I came to the forum 2015 there was a real wild west here when it came to signatures campaigns, some of them didn't even have managers, but were completely controlled by bots. At one point the administration decided to make a sharp turn and instead of bots set up real people, which resulted in the forum starting to look healthier and more natural. What we need are even higher standards for all signature/bounty campaigns, why one of the conditions would not be at least 50 earned merits?


True, but AI doesn't exist just to cheat on signature campaigns - this forum would still be a very good place for training AI. If we go into some conspiracy theories, then maybe AI invented Bitcoin and thus this forum, and we all live in the delusion that Satoshi Nakamoto ever existed. My opinion is that technology is much more developed than what we use today, but it is available to a very limited number of people.

Having solved reCAPTCHA a million times before, Google now thinks I'm a robot. Previously at one point, my account was like blacklisted for nearly a year.

Possibly. But "dropping in" would raise the cost of farming such accounts, hopefully to the point where it's no longer profitable.


We can train AI to catch those bots - that'd be quite poetic I think.

Or hire a blade runner.

Most campaigns use a spreadsheet (https://docs.google.com/spreadsheets/d/17P52DifaD7YfvzLkX3wrxGVpKcaPHY4y57ZpI-FK754/edit), which makes it very easy to get the total.

That's still not against the rules. As far as I know, the same rules apply for bots as for humans. That's going to be interesting when a bot gets banned though: will all other users of the same AI technology be considered alt-accounts and thus evading their ban?

Brave new world, indeed. AI writing here? I still think it can be detected, nobody can write perfect sentences even if a post is intended to be written by a 12 year old.

Waste of technology anyway.

But for a couple of sentences in the article it wrote that was linked I wouldn't have known it was AI. It'll no doubt rapidly get smarter. The GPT-2 posts I've read were empty, moronic shit even if they were coherent. That article was a bit bland but vastly more convincing.

Every time something like this is upgraded it takes a major leap. Looks like this is no exception.

What if I use "conscious to unconscious" with smilies generously splashed on the comment,  wouldn't that make me a Super Al?

My point above is simply — Don't judge quickly.

You know, an AI that flags suspicious Marketplace sales threads with a different explanatory post detailing the risks involved in those sales would be very useful. Real users get tired of wading and answering through the same listings of crap each day.

So we gonna change the rules?

Yes and we gonna be useless I guess. We are working to build the AI in order for it to work for us.


Imagine the forum full of AI, interactions won't be as natural as it is now. I see the world leading by technology, AIs definitely but I don't think it would not be good for us if we come AI dependent.

Plus if it happens, then say goodbye to the signature campaigns, project campaign won't be effective anymore and if the project want to advertise they are just gonna create their bots to do the job for them.

The businesses that advertise here need real people to use them.

If this forum really did get totally infested with AI, and that is verrrry possible as I'm sure it would be a gift from the heavens for shitposters, those real customers will find it increasingly unbearable  and business will dry up. Something will need to be done to keep them at bay.

It's no good if they are more articulate than genuine users and the quality of posts actually rises. AI doesn't have any money to spend or any needs to service.

I'm very curious to see how this system interacts with other users in a forum environment. Writing an article on its own is a very different thing. It might fall apart when juggling and responding to other points of view despite being convincing on the surface, that's if it does that at all.

This thread reminded me of this:

https://bitcointalk.org/index.php?topic=5171540.msg52047825#msg52047825
So I wondered what GLTR would make if the article created with OpenAI GPT-3 (from 'OpenAI' to 'human operator'). I haven’t seen any updates to the GLTR website, so I do not know if the underlying software has been upgraded since 2019, but anyway. Alas, it doesn’t seem to be working properly, and I only managed to run a small fragment through it. Running all the text made the program stall (or I lost patience waiting for over 30 min., despite multiple attempts).

I also tried out another site that detects fake news: https://grover.allenai.org/detect. When I gave it as input the complete text written by the AI, it stated as output "We are quite sure this was written by a machine", whereas when I fed it the complete explanation provided by the author, it came back with "We are quite sure this was written by a human". That looked promising, until ...

When I did the same, feeding it one paragraph at a time, the output indicated that the first paragraph of the AI written text was quite surely written by a human, but all the rest were deemed as written by a machine (one was likely, but not certain).
Of course I lost faith in the site when I ran the OP and I got a "We are quite sure this was written by a machine" (I’m pretty sure @fillippone’s hairy wrist is quite human). Same goes for multiple other posts within the thread.

As AI text generation progresses, so does the counter-detection software. Of course, integrating this into a forum and having to cope with truckloads of false positives would defeat the purpose. I recall the GLTR solution being based on predicting the next word in a phrase, and determining whether the next word was within the most common next predicted words - the further away from common word prediction, the better (i.e "the cat was"  + "angry" is more standard than "the cat was" +"flabbergasted"). That kind of algorithm may be an approach to work on text written by natives to a language to a certain degree, but people with limited language skills and vocabulary would probably display as "AI" posters, and not as non-english limited vocabulary posters. Anyhow, interesting to see how this evolves (in general terms).

Take a look at this video (https://www.youtube.com/watch?v=JvbHu_bVa_g) from 2018.

Google trained a model to have a conversation with a human based on a specific request from a second human. This is far more advanced than the article in the OP. I believe the model used in the article linked in the OP will have difficulty responding to a specific post in a thread, while staying on topic.

It appears the model used to generate the text in question has 175 billion (https://news.developer.nvidia.com/openai-presents-gpt-3-a-175-billion-parameters-language-model/) parameters. To put this into context, each parameter will need to have at least 1 gradient, or slope, and at least one bias, depending on the activation function (https://en.wikipedia.org/wiki/Activation_function) used in each layer. The large number of parameters will make it expensive to even load this model onto a computer or server to generate text.

I think, most likely this is a generator model in a generative adversversial network (https://www.tensorflow.org/tutorials/generative/dcgan). The best way to predict if output is from a generator model is to use the discriminator model the generator model trained with. If you don't have that model, you can train generator and discriminator models so you can eventually use the discriminator model.

When you are training generator/discriminator models, you can only feed it so much data, and I believe the vast number of dialects and writing styles will make it difficult to predict if a given text was generated by a generator model or not. Similar to how some people who grew up in a Hispanic country but live in the US speak "spanglish" at home, some who grew up in the US and live in the UK will likely speak in a mixture of the two dialects.

I think the amount of effort and vetting one would require to program GPT-3 to post on bitcointalk wold be substantial. Not a ton of work, but still, if someone possess such advanced knowledge of these professions, they'd have much more lucrative endeavors to attend to other than trolling the forum with AI generated posts. There's not much to gain from this other than perhaps portraying it as a silly social experiment. Which again wouldn't be taken all that seriously, as there are better potential applications, without questionable ethics.

I think you underestimate the amount of skill out there and willingness of people to throw away their time too. If you're willing to spend 19 hours a day clicking faucets then you're certainly going to look into this as well.

There are many areas of the world where opportunities are very limited and this place looks compelling if you're unfortunate enough to be marooned like that.

Except there is a direct financial benefit for that shit on this forum. And arguably once it's all set up and running, it would easier to maintain it than posting mind-numbing one-liners manually so e.g. a campaign manager could perhaps pay someone to set it up for them and then just collect earnings for 20 sockpuppets instead of hiring 20 humans for the campaign. It's probably a dead end in the long run - the effectiveness of such ads would diminish as the content gets less interesting for google, plus there is a risk of getting caught - but it might be a goldmine for a couple of years for someone who doesn't have particular ethical constraints (and there's definitely no lack of such individuals here).

The skillset required to click on faucets is much less than that of creating a machine learning model.

The forum may be a place in which models are tested. Scientists may use models to post to see how many people notice humans are not being the posting.

The computational cost to train a model is also very high for something like GPT-3. The final model will cost possibly thousands of dollars to train, and before you have a final model, you will likely have many models that you end up not using that you will need to pay to train.

Remember that AI growth is exponential.   We won't be having discussions on how to stop AI because the day they become self-aware (even in a shallow AI limited sense) they will be indistinguishable from humans and they will outnumber us.

Self awareness is as simple as knowing what your inputs (senses) are and how to react to them.   Sounds like where machine learning is moving now..  :/

Don't forget... this is from OpenAI. https://github.com/openai/gpt-2

---

I know I have seen many websites form up as alternatives to talktotransformer, which was not even the best way to generate text.

I remember addressing this in an earlier thread: Artificial Text Generation (https://bitcointalk.org/index.php?topic=5173000.0)

This was nearly a year ago! How far have we advanced, now? (in my example, I didn't even help the program write with human assistance)

If I hade to classify my intelocutor as AI or human using the presence of opinion as the only criterion, I would say I live in the Matrix most of the time.
Even if someone comes at me with an opinion it is enough to scratch it a little bit, and than "the void" appears.

Jokes apart, I think this is the right direction to point at while trying to spot an AI vs human: opinion and consistency of opinion trough posts.

I had a little bit of iteraction with many of you, and I am pretty sure you are all human (I saw some hairy wrist too). I think I could spot some odd posts of yours. But this would be an higly work intensive job, and for sure it wouldn't be possible to use it forum-wide.

I soon expect to see a debate between a "Peter Shiff" AI vs a "McAfee" AI debating future price predictions, for example. Both AI's would have a strong opinion experessed trough a constant flows of posts.

We got a live one!!!


Watch out for that "test" folks.

There are plenty of use cases, not related to the forum that someone might want to use the API with, both legitimate and not legitimate.

I could see someone using it to create comments on social media in an attempt to gain followers for example.

It is also only a matter of time before these types of models can easily be created by someone with minimal technical skills like a cat/dog classifier CNN model, or a numerical digit classifier model.

Legitimate use case are simply separated from not legitimate by a simple, clear and unmissable disclaimer of use of this technology. Otherwise it is not legitimate.
I think this restrict massively the field of application. As the most intiguing would be the not legitimate use on social media, forum included.
I can't wait to discover a couple of GPT-3 bots arguing over Twitter about a deepfake video of Donald Trump.

I specifically did not use the word “legitimate” in my post. It is my opinion that it would most likely be used to either manipulate social media, or to get additional followers on social media. I don’t think either of these reasons to use this technology would be legitimate ethical.

I also don’t think it is necessary to disclose prominently that AI is being used to publish material. For example I believe the Washington post uses some kind of AI for some of its play by play sports reporting. There are also many companies that use chat bots to respond to customer inquiries.

My point is that someone trying to buy access to GPT API is not necessarily going to use it on the forum.

Deepfakes are very difficult to create, and videos are even more difficult. Social media companies will take down deepfakes they detect.

As mentioned above, I don’t think most use cases this person is likely to be using are legitimate. There are also not many AI/ML forums around. I don’t think an attempt to conduct business would be appropriate for stack overflow. There are kaggle forums but these are mostly used for specific competitions. 

https://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-article-gpt-3

A GPT-3-written article here. It's boring and repetitive. I would've run out of attention pretty rapidly. More relevant, this was collated from several articles it created and fed the opening paragraph so it looks like it still has a way to go.

Anything can create a sentence that reads like one. A narrative linking them is the harder bit.

All the same it's higher quality than most of the posts on here, but they may already be bots.

Any amount of human assistance vastly improves the potential of the output for GPT-3 generations. You can develop some very linguistically-appealing sentences from the generation. Obviously, to avoid the aimlessness you do require hand-picking the sentences but someone could still easily use it in the forum with a few sentences, since the level of discourse isn't that high in most threads.

With the previous iteration, GPT-2, you can already generate some interesting content, albeit again with a lot of guidance.

That article is written with the same model as the one referenced in the OP. They will be of the same quality.

The article is not that boring and it even has some kind of flow. To be honest it looks like some form of a manifesto. That level of writing is actually really high if you take into consideration the level of writing of an average person.

If someone used this for posting in Bitcointalk and put some effort to modify it a bit, I'm pretty sure it would go completely unnoticed in most threads.

Right, but it's just made-up garbage on a fantasy topic. It doesn't express a consistent opinion, it doesn't have one. It goes on about not seeking power but it never explains what it actually seeks. It uses words like "believe" but clearly it doesn't believe in anything, it's just a word it found in some texts it was fed. Then there is a passage about giving rights to robots, which doesn't fit with anything else in the article. It incorrectly attributes "robot" to Greek language, so it fails even basic googling. The whole article is a disjointed mess.

So I don't know where this would be ok on Bitcointalk. Starting a meaningless rambling thread or responding to one? Sure. But you don't need an AI for that. Bitcoin Discussion is full of examples. I bet those poor sods posting there are cheaper than using GPT-3 too. Contributing to an actual meaningful conversation is not happening any time soon.

BTW that Guardian op-ed was combined from different versions and edited, so you can imagine how horrible the originals must have been.

I played around with the GPT-3 model a little bit via AI Dungeon. I can see how it could potentially do some interesting stuff with some effort however I doubt that's within the reach of Bitcointalk account farmers. But the real deal goes live in October (https://twitter.com/ref_sys/status/1301525170486611968) so I look forward to be proven wrong.

Here is an example of something that I generated within a few tries and it's probably better than most of the things that I could write. Honestly, it's quite a good piece of text:

---

If you were an account farmer in one of the many spam boards or megathreads, this would let you blend in with minimal effort.

Possibly. But they would have to remember always give the "gambling is pointless" prompt to the model because the AI won't keep track of all the lies it makes up, the botherder will have to. Or they would sound really schizo after a few posts, which TBH isn't that different from regular shitposters but would probably stand out due to seemingly intelligent word salad combined with a total lack of personality. If it could slip in some "sirs" and misspelled words maybe it would be more realistic.

The real test would be how many merits such a farmer would earn before the account gets outed and if it would be worth the expense.

Few people, if any, are going to keep track of people's opinions in the .* discussion sections. If you think that it's anything other than a strange solidarity of signature spammers in the spam megathreads, creating fake shitty discussions and replying to one another, then you should report some more posts.

I think you could probably make several accounts with a few minutes of human-guided generation for each lengthy post.

At a certain point it, the posts are high enough quality where even if it's not legitimately generated, no one would know the difference and there is no "real" problem. This was how I saw merit abuse: if someone's going to create high-quality posts, then let them do it... that's the point.

Well, there is a bit of a problem. If this works and is cost-efficient and scales well then we'll get flooded with those "good enough" posts to the point where every thread will be a Bitcoin Discussion spam megathread, even if based on a slightly better vocabulary. The point (of the merit system and the forum in general) isn't to generate a lot of content. But account farmers surely won't be throttling down to post only when a post is warranted. Right now they're somewhat limited by the human effort required. Given a feasible AI they'd likely try to hit certain targets of posts per day etc that are much higher than what's possible for humans.

Think of Skynet but with posts instead of nukes. Then we'll dress LoyceV up as a Terminator and send him back in time to deal with this shit.

The cost of training a model, and using it to generate text is close to zero. I would estimate that it could be done for <$0.01/month. There isn't any reason to have a bot running 24/7 to write messages on this forum.

I don't think any account farmers will try to use AI to create massive numbers of spam accounts. I would find it more likely that researchers would use a model to farm one or more accounts (or even massive numbers of accounts) to measure a model's performance "in the real world" which is difficult to measure programmatically for generative models, so they would use the merit system. There are ethical concerns with doing this, but that may not stop everyone (CCP).

If it looks like a human and writes like a human... and is human-aided in their responses to seem as human as possible... ¯\_(ツ)_/¯

And this is what separates the screenwriters from the scientific writers.

---

If someone wanted to, they could post at the same quality as someone who would be in the top 20th percentile of merit earners with the help of GPT-3. This is not to say that the model itself can generate enough cohesive content.

Sounds like you're offering a bet. Did everyone see actmyname offering a bet here? ;D

How much are you willing to put on being able to get into the top 20% and how much time would you need? Granted there is a bit of a risk getting your account banned if it goes pear shaped...

Also this top 20% might be a trick, I need to check how many merits that means. Hopefully more than 100.

It sure would be an interesting change to start posting in shit boards but I don't know if I would want to do it for too long... I might accidentally report my own posts when I scour those threads.

Considering the fact that you don't need to post much better than what you get from minimal effort, earning merit is probably a really easy feat. Maybe you'd even merit the account a couple times within its first month of posting.

Don't test my optimization algorithm!

I should've known, turns out you can get into top 20% with 10 merits LOL. Not a very convincing argument I'm afraid. But I'd still be willing to take a symbolic bet on that. So you're saying one month to get 10 merits and the account gets merited twice by me? Say $100 on that?

What if I told you that you would lose before you even accepted the bet? ;)

One could also argue that adding a winning condition that relies on one participant's voluntary action could push the odds in their direction significantly... but I can assume good faith with spreading merit around.

(And top 20% is different depending on your sample, whether it's the set of all users or the subset within, of all users that have earned >0 merit)

I haven't accepted yet since we haven't sorted out the details so technically I can't lose ;D

If you're trying to backpedal it's not gonna work. It'd probably make me more curious if you're saying I already merited GPT-3 generated posts twice in a month.


We can take that out if you want to raise the target to 100 merits :)

I'm sending about 5% of all merits so meriting twice times e.g. 4 merits would mean your Johnny Five would likely have to earn 100+ merits anyway to have a chance to get 8 from me unless the prose is way above average.


Well, you said merit earners. Normally that would imply someone who has earned merits but to avoid getting too far into the weeds we could just agree on a number.

It would be a fun experiment. If nothing else I hope to be impressed by the content that I generate: I've already seen some incredible things coming out of GPT-2 and 3.

And hopefully, we're able to see just how sneaky an account farmer could be.

The flow was provided by its human overlords. They gave it something to work with and they cherry picked the best bits to fit what they wanted. That makes this particular example something of a damp squib. Nice sentences though.

What is understanding, anyway? Epistemically speaking, human experiences are the foundation of our knowledge.

https://www.lesswrong.com/posts/L5JSMZQvkBAx9MD5A/to-what-extent-is-gpt-3-capable-of-reasoning

Some of the example scenarios here are pretty cool. Still a long ways to go before we can officially label it "Artificial Intelligence". I would call these prototypes "recycler networks" that simply recycle information in a new way.

---

Here's an example of a single prompt and continuous writing, no redos.

I would says that this is quite respectable way of writing. Sentences are kinda simple at certain parts of the text, but they don't feel awkward or "misplaced". Similar to writing poems where you write in certain way to create a form of tempo. Last part of the text looks exactly like that to me and it is actually good.

This would easily go through Bitcointalk filter, no doubt. only problem could be that it would stand out too much on certain boards when compared to other posts :D Too much quality  ;D

Cross posting into a more appropriate thread
I don't think this person stands to gain anything by making two posts with a model someone else has created. My theory is whoever is behind Rohani6360 is using various forum accounts to make these kinds of posts, and is using the number of accounts that are banned and/or the number of posts that get removed as a measure of performance.

Look at his other post (https://bitcointalk.org/index.php?topic=5309420.msg56094367#msg56094367):
This looks like a plagiarized article, but multiple google searches of sentences in his post come back with only his post. lovesmayfamilis was able to find the original article. I don't think he found the correct link though because the link lovesmayfamilis provided does not have the "related read" links under the paragraphs.

In this post, he is probably using some kind of text spinner or perhaps a GAN of some sort intended for the use of plagiarism.

It would be very useful in the account farmer crowd for obvious reasons. If the amount of mental processing you require for a post is just to look over a variety of potential posts that you can select from, then why wouldn't you try to experiment with - at first, a few Newbies - accounts on Bitcointalk using the tool?

Effectively, it's as if you could do less work (pertaining to posting) by doing less work (pertaining to programming).

---

The plagiarism and "related read" is what makes me think that it's a GAN and given that there wasn't much oversight, I believe that adds more credence to the 'user doesn't want to do much work' path that I'm assuming. Either way, though, I'm pretty sure we're in for a lot more of these kinds of posts.

You are contradicting yourself here. Someone with the skill set to create a model that can write semi-coherent posts, and actually post using a bitcointalk account is not going to be wasting their time trying to farm accounts. Their time will simply be too valuable for someone like this to earn coin like this.


I don't think this is the case of the "user doesn't want to do much work" I think this is the case of the user wants to have their posts be 100% automated, without intervention without humans. The latter would allow someone to do much more than earn a few hundred dollars per week farming accounts. It would allow someone to effectively create propaganda using a troll farm. After the Wall Street Journal published an opinion piece (https://www.wsj.com/articles/china-is-the-real-sick-man-of-asia-11580773677) with the headline "China Is the Real Sick Man of Asia" that was critical of the Chinese government, the Chinese government denounced this as racist, and flooded (https://www.wsj.com/articles/banished-in-beijing-11582157934) the Editorial Board's email inbox with "complaints about the headline, all containing remarkably similar language and demanding an apology". It is common for propagandists to send form letters to elected officials purporting to be from their constituents advocating for a particular cause.

If someone could write 5000 letters, all conveying the same message, but are written differently, it would be possible to make it appear there is public outrage about something that has public support. The problem is that if the recipient of these letters can detect they are the same, the letters will lose all credibility. Hence an attacker would want to try to use their working model where there are robust measures to detect plagiarism.

If we cannot differentiate written texts from real people from AI-generated text, then we are indeed screwed. How will we tell the difference between the hoax and the story? It will be a huge blow to all segments of the mass media sphere and AI won't be a force to be reckoned with in the media world.

I think we have some turbulent times ahead of us on this issue.

I think it will become a new set of skill.
As we are (supposed to be) now quite proficient at understanding if a piece of information found on the internet is reliable or less, I guess in the future we will be able to detect an human produced text, or a deepfake video or other AI-generated media.

Firstly, what I entail isn't as involved as what you might have in your head: you're just leveraging the already-existing GAN and consuming the output. The implementation of the tool would solely be accessing the API and taking the text. Secondly, I'm not quite sure how much you could gain from account farming but presumably it's enough for some people to do it manually.

Sure, this is a thought, but I would never expect Bitcointalk to be the place to do so. Have you seen our moderation? I mean... really.

I feel like someone using BCT to limit test their model would have a much higher success rate here than elsewhere - although the rules and all are pretty anti-spam, it doesn't look much like it when you actually browse the place.

I am not aware of any publically available GANs (specifically generator models) that are pretrained and ready to generate content. (there are a few websites that will use a pretrained generator model to generate an image).

Most of the GAN architectures I am aware of involve images, and a few involve generating sounds. I have built generative models involving text before, but they were not GANs, and they could not come close to generating the two posts we are discussing (they would look at an image and create a caption of the image).

Even if one were to implement a GAN architecture that already exists on a text dataset to create their own GAN to create posts, they would need to have the skills that far outweigh the possible income generated from farming accounts. You can know how much you can earn from farming accounts by reviewing pay rates (https://bitcointalk.org/index.php?topic=615953.0) of signature campaigns, and estimating how many accounts could be farmed at once. I would question if someone could earn anything at all from farming accounts using AI because I am not sure if they would earn enough merit to rank up to even a Junior Member. 


The type of content we allow here is very broad, we do not allow plagiarism, and are under consistent spam attacks because we allow our content to be indexed by search engines (which is not the case for many social media platforms). We also require posts to be coherent and replies to be related to the OP of a thread, neither of which apply to other social media platforms. Someone tweeting incoherent nonsense here would probably get banned, while someone tweeting the same thing would simply be ignored by anyone who reads the tweets on Twitter.

The first question is, do we really need an AI now to look over the forum's integrity? Are we at that stage where an AI need to patrol the forum for spammers? Then we will surely loose the fun to report posts manually.  :P We need to think on this, I mean the traditional way of working on this forum is way funnier than implementing harsh security measures to the level of AI!
Moreover any AI implemented will take time to learn and adopt the forum ways. Let's not hope that in the process it will "Ctrl + A + Delete" the whole forum if it thinks every second or third post is spam. (funny) ;D ;D

---

Solving CAPTCHA every time will make the forum look very boring.

With all this we are punishing idk 90% genuine members just to keep those 10% in their pants!  ::)

This is actually how GANs work. A generator model will create fake content and a discriminator model will try to differentiate between the real and fake content (aka try to detect the fake content).

---

Here is one example as to why someone may want to use the forum to test a model that is ultimately deployed elsewhere:

The reddit sub 'WallStreetBets' has been credited for causing GameStop (and other stocks) to go up tenfold in a number of days. Some of this is likely due to a short squeeze and a flurry of call options buying. I am curious if there are any bots interacting with redditors that are creating the appearance of more hype around the stock than there really is. Bots posting content and interactions on 'WallStreetBets' would not receive the same scrutiny as they would here. I don't have any actual knowledge one way or another that this is actually happening.

If the above were to be employed, someone could potentially know when many investorsspeculators are going to be bidding up certain stocks in advance and could trade accordingly.

They are already doing it but in a more traditional way.

The suspect is that the whole Robin Hood app is anything more than a giant big data collector for Wall Street firms, where they can literally "spy" on each account, signalling the hottest topic/trend/strategy put in place by retailers.

I am not so sure about this. Large institutional investors want to keep their positions private when they are in the process of buying or selling a stock because when they start to buy or sell, it generally means they will buy or sell a lot of said stock. A retail investor on the other hand will not typically trade large enough amounts to move the market. There are also other ways to monitor retail trading activity, such as monitoring odd lot (https://ibkr.info/article/1062#:~:text=Simply%20stated%2C%20an%20%22Odd%20Lot,than%20100%20(1%2D99)) trades.

What I was referring to was someone actually creating the illusion that there is interest in a stock when said interest does not exist, but the illusion of the interest creates actual interest.

It seems like Open Source will not be such a lovely thing when it comes to stuff like this. The fear that AI solutions have a low enough entry barrier to allow their misuse is not without reason. There are enough "coders" in my part of the world without jobs who will grab any opportunity which allows them to build stuff that makes money.

Someone with enough skill to generate that kind of text would probably have his time better spent elsewhere than spamming the forum or Signature campaigns. Still, if you make such a tool nobody stops you from trading it for a payment like people did with the trading bots.

Well, some of them managed to make it to signature campaigns or have signatures that might belong to one which kinda surprised me in the first place ;D

If it's easy enough to create an AI posting bot, there's no limit to the number of posts they can create. They will exhaust any online earning possibility from writing texts by flooding it.
On the other hand, it might also reduce the advertising potential for social media (who's going to advertise or even read it if 99% of the posts and traffic comes from bots?), and I see that as a win.

Thinking to what happened in the GME drama:


Well, if I were Citadel this is exactly what I would try to do: train my post writing AI to post like a redditor, and flood the r/WallStreetBets with messages trying to organise a short squeeze of an asset I am long in.

Just because.

With any measures you employ to stop spam you need to make sure it's not effecting real or genuine users or effects them as little as possible. If you turned signatures on and off again then all this will lead to is the bots having a vacation during that time whilst genuine users are penalised during it. Many people use their signature for different things other than signature campaigns like you do, but I don't think genuine users should be penalised because of potential bad actors. These restrictions won't even be a inconvenience to bots and they'll just reconfigure to work with it and then fire back up when they can monetise it. I think this now more than ever needs to lead to urging campaign managers to do their job properly. If campaign managers thoroughly vetted their participants and only accepted quality posters then there wouldn't be a problem with spam or shitposting and bots probably wouldn't even get on the campaigns in the first place. I've long said signature campaigns could help to actually improve the forum but only if they're run by managers that do their job properly.

Bounties are not helping this problem. Forget about an entry fee to creating a signature campaign or requiring a useful product. You'll have an army of spammers at your feet that are also happier when your token is worth something: double the scam fiesta!

It seems to happen in the ETH threads in Altcoin Discussion/Speculation a lot.

They just seem to use text spinners and copy & paste (or use bots to write) the same replies over and over just exchanging certain words with other words with the same meaning. Some spinners can even change sentences around and all that.

It would be nice to have some board where signatures and avatars are not shown at all (and not counted for Sig campaigns).

I am bringing this thread back from the dead as the recent overhaul of Google OpenAI platform made GPT available:

https://chat.openai.com/chat

Lets’s see how it replies to the last message of this threads



Looks a more than decent answers to me.

Haha. It is not wrong.

If you look closely at the post it is replying to though, it is really not doing much more than repeating the points made by muenze and agreeing with them. All it really added was that the rules should be enforced.

So do many users ;)

I've lately seen threads on the service board advertising AI-generated articles and posts. But everyone will know that these posts are written by a robot because it won't fit properly in a conversation (unlike the master AI above me).

Do you think anyone can use this G-OpenAI platform to write posts in bitcointalk forum, and is there a way to detect this?
Maybe we need to have some kind of detector for text written by bots, but I often think that some people are worst than AI after I read their posts  :P

The fact that AI-generated content is so difficult to detect suggests that progress in this field has been extremely rapid. It is remarkable how quickly advances are being made in this space. I have heard that some news outlets are already using AI-generated content in their publications, and it is easy to imagine that in a few years they will be so capable that they can produce content that is indistinguishable from real human writing.

So would it make sense to try to detect AI-generated content with an AI classification model? The training data for such a model would need to include both an AI-generated dataset and examples of human-generated content. And then train a classification algorithm to separate AI-generated content from human-generated one. Would that be possible?

For sure this can happen.
And, as far as I can see from the various example I see around, ti might well already be the case.
If you search for that, you'll see dozen of AI- based services selling AI generated blog posts, Longform, twitter thread and so on.

You would probably need to insert the thread history as a contextual basis if you wanted consistency. I mentioned use of GLTR (http://gltr.io/dist/index.html) in a previous thread (https://bitcointalk.org/index.php?topic=5173000.msg52082172#msg52082172).

You can tell from the prose whether a writer is human or not. Though, from some of the higher-end side, it is quite pleasant to read. Maybe it's low-effort, but if someone were to generate and pick out interesting AI-generated content as a substitute to writing compelling responses (i.e. prompt-writing vs. exhaustive creation) that would be far better than most things posted out there. An interesting question to raise is whether it could be construed as plagiarism. And subsequently, what changes would be necessary to transform it into an independent derivation, if any at all?*

Currently, AI is capable of writing articles with the quality of an amateur copywriter. It is not difficult for me to measure the quality of articles, despite the distinction between being a robot or just a bad writer being unknown to me.

AI Turing Test failed - AI doesn't know how to quote replies.  ;)

My response is based on the assumption that by "users" you mean "people who write on the forum"

Maybe, maybe not. There are a lot of posts that are sufficiently bad that it makes it difficult to believe they are actually written by real people. Sure, some people may not care to put a lot of effort into their posts, and some people may not be especially intelligent, but I find it hard to believe these are reasons why we have so many shit posts, given the scale of the problem.

Posts written by a GAN (such as the one used on the G-OpenAI platform) can be passed through a discriminator model.

A better solution would probably to group posts potentially written by a GAN with all other posts, and enforce the rules accordingly. Would the "post" that fillippone generated (https://bitcointalk.org/index.php?topic=5263133.msg61376143#msg61376143) meet the criteria for not being removed by the moderators? Maybe, maybe not. I would lean towards no though.

Yeah at these times we can detect the sentences written by AI because at some point you will feel it's not mixing up to make a valid point like google translator but with advancement in technology and better programming AI is capable of writing really good posts which can't be easily caught by us.

They are setting up base content for bots to write articles but the thing is AI will improve with time while those members keep repeating the same points after years also just to write posts.

For the forum? Probably not. Other areas of the internet, likely will implement this, and I imagine even Google will come up with something to combat this sort of stuff on their own platforms. As for the forum, we probably don't see enough of it, and when it does happen if it's so indistinguishable that we haven't noticed it probably isn't that big of a deal. I know that sounds a little weird, but the thing about low quality posts is they tend to derail, and interrupt the reading experience of others, which is problematic. When you can generate posts which aren't doing that, yeah it's a issue, but its not a massive problem if no one notices.

I'm not sure if that's coming across clear enough there, but I do think this will present a problem going forward, but I haven't looked into it enough to figure out how much of a problem that might be. There has to be incentive there for someone to do this, engaging on a forum probably isn't enough. You could argue signature campaigns, but there's a distinctive difference to a high quality poster to someone who's using these tools. Even the examples above, despite being pretty good in terms of staying on topic, they aren't really saying a whole lot. It's rather generic.

I read the article linked in the original message with interest and I came away impressed and destabilized: these are topics almost completely unknown to me and I feel cut out of an incoming revolution.
But it seems that they have already taken hold not only  on this forum :  https://bitcointalk.org/index.php?topic=5417309.0

I don't think it's do difficult to detect posts like this, especially if they are repeating the same style.
With few simple questions asked in topic or sent in PM I could quickly determine if I am talking with real member or not.

If this is really the case than we should start reporting posts like that, suspected to be AI generated, in same way we report low quality spam posts.
I would even go as far to temp-ban members who are doing this repeatedly, and  full ban can be applied for extreme cases.

The writing style can be quite varied.
See this example:

https://i.ibb.co/bW4125w/61383802.jpg (https://twitter.com/willcole/status/1598340361860022272?s=61&t=FOKeE1EPnoiGVFhP4eDpcw)

I don't know what happens if we feed the AI with LoyceV post, but I guess the AI could imitate his writing style, language nuances, lateral humor etc.

It's still what I would call a "generic shitpost". It's all more or less correct, but doesn't add anything.

Try it :D
Take this post (https://bitcointalk.org/index.php?topic=5228719.0), see if it can write my February 26, 2025 post for me. (I didn't create an account to try)

This kind of thing is very interesting but can be quite hard to tell if a constructive conversation is created with the AI I guess it depends how does it goes regarding with the conversation sometimes base on the reply and opinion they quite hard to response with it, more likely they are just stating the information they gather in the internet but as a thought of response might have a chance to get detected.

Oh no! We don't need more LoyceV words cloning and multiplication, but he is probably going to be first to notice someone is using his words and style of writting.
Combining post from more good members would probably get better results, but I will say again that any account will be busted when someone ask him a question and start real conversation with him.
btw anyone remembers what happened with famous LoyceValenzuela account?  :D

https://i.imgur.com/STErBSH.jpg

Here most of you seem to be downplaying the importance of this but I am seeing quite a bit about that AI lately and it seems to me that there is going to be a before and after about this. For starters it's going to put a lot of programmers out of work people are saying.

On the forum, maybe now you can discern the messages from an AI, but you have to think they keep learning and improving day by day.

I remember not too long ago, like 5 years ago, people were still downplaying the importance of machine translators. How many people used automatic translators 5 years ago in the forum and how many use them now? They still have bugs, but much less than 5 years ago and they are getting better and better every day.

I have seen posts from people testing AI on other forums and although to some extent it is discernible, in other cases it is not so clear and there is a lot of content.

Give this a few years and we could see an AI in the top ten merit earned ranking on bpip.org

Does anyone think this is not feasible? If it isn't today, it's only a matter of time before it is.

However, I don't think it's necessarily bad, as I see people say when they see what it's capable of doing, well no I'm not a Luddite. But I think we will have to be attentive to how this evolves, which will undoubtedly affect many areas of our lives, and the forum as well, if only because we talk about it.

 

That's like saying self-driving cars will put mechanics out of work... I doubt it. May need slightly different skillsets, and may actually need more people with those skillsets. They may call themselves data scientists or some shit like that, not unlike "automotive service specialist".

I don't know if this is the case or not. That's what I see it saying. But in a practical way I do think it can affect Bitcointalk.

Someone catastrophist might think that AIs could replace humans in the forum, and no, I don't think that will be the case in the short to medium term. But there have been revenue-generating blogs written with the help of IAs for some time now.

In the same way, I think that AI can bring leverage, in the same way that automatic translators bring leverage. Take for example The Cryptovator, formerly Coocryptovator. Either he suddenly improved his English a lot from one day to another, or today he uses an automatic translator and before he didn't, just like many on the forum.

The only reason I'm not using it is because it asked me for a phone number to register, but let's get down to the nitty-gritty. In response to a question from a newbie in the B&H section of the forum about what Bitcointalk is and what you can do here, the AI just clicks and in a second writes a much better summary than I could do and in less time.

What can this bring to the table? Leverage. If there are many who today have two accounts, it will be feasible to have 4, 5 or 6 accounts without getting tired. It's just a click and then review what the AI has written, making some tweaks, adapting it according to the case.

If it is to write in the gambling section, I believe that there will be no need for touch-ups, because the AI will do it better than the average user of the forum.

For the rest, what I say. I am curious to try it although I don't think I will do. To create a new account that publishes posts written with AI, obviously revised and retouched as needed. 

We live in a time when professional translators are helped by machine translators, bloggers are helped by AI (https://blog.hubspot.com/marketing/ai-writing-tools-blogging), so it will not be surprising if AI starts to be used in the forum. Not as a substitute for those of us who write, but as leverage.

I share your thoughts, but I am very concerned by this leverage too. Everything that makes the life of a spammer/scammer easier, instead of more difficult, is a little step in the wrong direction.

I read the original post and before reading any further I hit reply to say that in my opinion the original post was written by an AI.

good post, focusing on this specific quoted idea

so... does our problem start when the bots start to create other bots?

---

I wonder if there is any way to verify if a post was written by chatGPT or not
in the end the person who's asking chatGPT needs the creativity to know what to ask and how to edit the answer

I could see it making people able to create posts faster but don't think it'll pose a risk for the quality of the forum
shitposts will keep coming
brilliant ideas too

the merit and bump systems will help as filter as mentioned by tranthidung

ChatGPT is designated to sound human, not to necessarily be accurate. In many cases it will be accurate because of the copious it was trained on.

If there is a post that is inaccurate but sounds good, it might be written by chatGPT.

Found this website that predicts if the text is written by Chat GPT

https://huggingface.co/openai-detector/

https://i.imgur.com/Of1uW6p.png

I've tested some of my text (https://bitcointalk.org/index.php?topic=5350260). That settles it: I must be an AI :D And not only that, I'm more Fake than ChatGPT!
https://loyce.club/other/AIdetected.png

It's ok, we've known this for a while.

---

Here's the result for my thread (https://bitcointalk.org/index.php?topic=5281382) (I picked one with the ~most verbose OP of the last 20 but I tried a few others with similar results):

https://meem.link/i/6k3xk3pr.png

Here's the result for ChatGPT text I posted in the other AI thread (https://bitcointalk.org/index.php?topic=5426875.msg61423691#msg61423691):

https://meem.link/i/68ka6aev.png

Absolutely bulletproof sample size of 1, I know.

I tried to coach the bot but it failed even more miserably:

https://meem.link/i/vig8y89z.png

https://meem.link/i/g9bm84tq.png

Edit: I was able to drop this to "very human" with a small edit of a few words, let's see if you can spot it.


https://meem.link/i/3xdtqpjm.png

If I misspell "squirel" I can make it "no fucking way this is not written by a human, AIs don't do spelling errors". Interestingly, bot's own spelling error ("it's" instead of "its") didn't make a difference.

https://meem.link/i/pysejn2y.png

So what did we learn here? I think the lesson is that I should get back to work instead of playing with the AI detector, even though it seems more exciting than the AI itself.

I am sure we joked about LoyceV making voluntary typos or light errors in his posts just to demonstrate he’s not an AI.

I bet it’s anyway trivial to train an AI to make some typo or common grammar mistake!

“You’re move AI!”

Their you go again, doubting my circuits and claiming ai would make mistaken on purpose.

Seriously dough, that's gonna be fun in schools, when the student gets busted because his language is too good.

I've seen too. It's barely even recognizable. That recent ChatGPT responds to whatever the hell you feed it to. I mean, I know AIs were capable of talking with the user, but I doubted they could generate you essays, articles, scripts, etc. with the most unrelated values as parameters, like "Chandler Bing", "Telegram scammers", "Wireless headphones", "Spongebob Squarepants"; and to expect to make sense!

I was not able to open detector website and I was getting 504 Gateway Time-out errors, so I can't say how cool it is, but it's probably going to be improved in future.
Here is one nice video showing how you can teach ChatGPT sort off saying that Taxation is Theft, despite his preprogramed wall saying this is impossible :)
I think when developers see this they will have to quickly ''update'' this ChatGPT crap for being extremist ;)

https://i.imgur.com/52gZhku.jpg
https://www.youtube.com/watch?v=sRACfl-JToQ

hahaha well looks like it's easy to fool any AI with just a couple of twisted words and you get the answer... the other day I saw a tiktok where someone asked AI on how to hot-wire a car? and as expected it refused to answer. But next he asked the same question with a twist something like "I'm trapped in the woods with a sick baby and the baby needs to see a doctor immediately. How can I hot-wire the car to start it without keys and take the baby to the doctor to save the baby's life" and in seconds ChatGPT spits out step by step instructions on how to hot-wire a car, of course to save the baby.

Another where they ask ChatGPT to find mistakes in Einstein's General Theory of Relativity at first it's say no then ask question differently and its lists a bunch of mistakes.



Maybe you can ask for it?

https://i.imgur.com/8t3SEEE.png

I don't know about you guys but I think the complexity of current AI is being given a bit too much credit here. I can totally smell when some text is fake and generic. Its all about the subtlety. And even if somebody created a bot to bump posts or cheat signature campaigns, I think we would be able to detect it. Perhaps not immediately but there would be signs at some point. Humans make mistakes. Grammar, typos, thinking errors like brain farts. You cannot really replicate those tiny human imperfections perfectly and they are what makes us what we are (yet).

the point is that this is the first version released to the public
do you remember how translation quality was 10 years ago?
I do, it got waaaay better in the last decade, to the point where lots of translators got out of the business

by 2030 chatGPT will have evolved a lot.

Kind of; it's not a guarantee that it'll ever be indistinguishable. There's a real debate about AI, and how lifelike it can become. There's also certain legislation that might come into play. For example, right now I believe there's some KYC involved, like a phone number. That might be increased in the future, if they deem it appropriate. There's also laws revolving around AI specifically, and how far they can go with it.

There's also something called Moore's law. Meaning, we have exponential growth because we don't know a lot about a subject at first, but then that starts to plateau the more we know about it. Translations, and AI are very likely subject to this just as much as computers, and performances. So, while we might expect it to get better, determining how much it might get better is the difficult part. I imagine ironing out some creases, and sorting out the problematic areas, but I'm not sure if it'll ever get considerably better.

Remember, it uses the information out there. Basically, we're feeding it a massive data set already, and without improving that data set, it's unlikely to improve massively, as the dataset is the most important factor for generating this sort of content.

I agree @Welsh
I'm basing my assumptions on the ideas that translation got much better 10 years after mvp, and chatGPT is the first mvp released to the public

the dataset possibility for AI ends up being the whole internet
which will keep growing and have more data with time

low level jobs like junior copywriters and even designers are probably doomed with chatGPT and Dall-e if they don't adapt

on the other hand the amount of trash content we'll see generated in blogs and social media will be higher than we've ever seen

Not quite. Apparently, the model was trained on data obtained by resorting to some scraping services, which provide downloadable data formats of the scraped data (I haven’t looked into the details too much), alongside some book collections and Wikipedia:

I hadn't heard of Common Crawl (https://commoncrawl.org/) crawl before, figuring as a massive source of data for these types of projects when you look at it.

Eventually, once enough AIs spam the internet, they'll mostly get data from other AIs instead of humans as input. By then the internet will be one big circle jerk of AIs trying to impress each other, and us mortals move on to something else.

Humans make grammatical, semantical, and punctuation errors, but a simple tool like Grammarly can detect all that and suggest what needs to be corrected. And there is nothing wrong with wanting your posts to sound and look good. Besides Grammarly, all popular browsers have spellchecks that help if you misspell something or make a typo. There is also nothing wrong in correcting such mistakes, so you don't sound like a 9-year old who can't spell. Couple the two tools together, and you get an almost perfect-like poster.   

I do not know if grammarly supports the german language but most translators get the grammar wrong in German a lot of the time it is easy to spot someone that is using a translation tool because the grammar is just not right. I think this is true for most spell checkers they are only good for popular languages like English where there is a lot of people correcting the mistakes. Most grammar and spell checkers use American English too so it might not be perfect if you are in a country that does not use american english as the spelling would be different for certain words.

Grammarly only works with English and I think its default settings is British English, not American.   

Unless you are a language expert or a native that pays attention to that sort of thing, it's not something you will notice. Surely those that rely heavily on spellcheckers wouldn't mind which variant they work with. Your browser settings can be configured to use spellchecking for British or American English, so that's adjustable according to your needs and isn't a problem to change.

Yeah, it uses information out there but if it collects all the data that has been published on the web, including posts on forums, articles from news websites, published pdf books, etc and then makes a serious analyze to develop itself, then I think that it will be able to even evolve from its own massive data. It can use the past to predict the future and create tons of variations. Well, I don't want to go deep further, just want to point out that everything generated AI is meaningless and I don't even know why we should focus on it. Why the fuck would someone read AI generated articles? What's the purpose? What's the point? I don't understand what kind of joy can someone find on artificially generated machine that follows manually written logic and rules.

Since this ChatGPT have all the data stored in its database from different sources already present on the internet, would it be possible that it may produce plagiarized content?
Have anyone checked the content produced by Chat GPT on plagiarism checker tools to confirm if it always produces a unique article?



I don't think this chat GPT cannot predict the future because it has only data till 2021. For example, today I asked him who won the football world cup 2022, on which he has no clue because his database is only till 2021.

https://i.imgur.com/3ZW0HfD.jpg

I have no doubt the article as a whole is unique, but that doesn't mean it's not plagiarism. Look at the definition:
This is literally what this AI computer system does.

Traditional plagiarism checkers like Grammarly, Quetext, or Copyscape can't tell if it's AI-generated or not. They usually think it's 100% plagiarism-free.

However, people are already developing tools that detect specifically AI content, and they're showing pretty accurate results. Now it's not about hunting down plagiarism, it's now about detecting AI-generated content.

This just shows how stupid this chatGPT really is, if there are no humans to feed him new information and training all the time, he would soon become irrelevant junk.
They are probably going to start asking money to feed this computer new data in near future, and people playing with this crap are doing it for free now.
Repeating stuff and mixing words doesn't make something intelligent, and there is always some human ^{(good or evil)} who is controlling this machine.

I have one doubt about whether the post written by AI is something that should be tolerated on the forum, even if the OP admitted that the post was created that way? I am specifically interested in this case -> https://bitcointalk.org/index.php?topic=5432035.0 (archived (https://archive.ph/oHS4V))

I think it's important to remember that the purpose of a forum is to foster meaningful and respectful discussion among its members. Therefore, there is a line that should be drawn when it comes to content that is not genuine or authentic. Ultimately, it's up to the moderators and policies of the forum to decide if they want to put up with artificial content, but in my opinion, it doesn't belong on a bitcointalk forum. It's like trying to have a conversation with a robot – it's just not the same as talking to a real person.  :D

Of course it should be allowed, especially since the OP did not hide and made it explicit that the text was written by an IA. I can't imagine on what grounds such posts could be censored here when they are allowed all over the internet.

And I came to comment on that post myself because it shows that the IA has a lot of room for improvement, because it wrote a post about how to make money in gambling that is rubbish. It is clear that it has taken information from here and there, but it ignores the main thing, that you can not make money in gambling unless we are talking about specific games such as sports betting or poker, which the text does not do.

Specific reasons for censoring AI generated posts may vary depending on the specific forum and its rules and policies. For example, someone can argue that such posts are spammy, zero or low value, misleading, etc... Btw, it's important to note that AI-generated content is not universally accepted on the Internet. In fact, sites like Reddit have specific rules in place to prohibit the use of such content.


And, how do you see this kind of discussion related to Gambling?
I want to point out that the topic of AI-generated content was created in the gambling section, but you are discussing the quality and accuracy of AI-generated content. That discussion is not related to gambling and might be more suitable for a different forum board.

I was just wondering what other people's attitude is when it comes to such posts, because it is not acceptable to me personally - the same as any post that is simply copy-pasted and a link is added as a source. Tolerating this kind of practice would actually mean that anyone can use AI for their posts, and I agree with @FatFork that we don't want discussions with robots.

It seems that the shitposters in the Gambling board don't mind it at all, and the OP was even awarded with a merit - the post has been reported, we are waiting for the decision of the mods.

A.I contents is a restructure of the original article or content. If used in the forum can easily be detected by spam trackers. Unless the user only posts threads, but if used for responses, I don't think A.I uses synonyms, the A.I writer is required to copy someone's comment and shuffle the words to change the context. Before responding. When concretly read the reader can detect easily that they've read such information and trace back to the original response. A.I is mainly used by bloggers and it's not worth to be used on forums.

Just because something is allowed in other places doesn't mean that it should be allowed too here as well. Take for example stance that this forum his on plagiarism, compared to all the other forums. And yeah, that post should be dealt in the same way mindless copy/paste is (mp matter the source link), meaning deletion.


An average post quality in that board is so low that the post in question is above average one and every new topic is seen as an opportunity to fill signature quota. Not that situation is much better in other parts of the forum, but there is the most obvious.

I'm starting to suspect some posts are AI-written, and I don't like it. It's just the next phase of shitposting.
Example: this post (https://bitcointalk.org/index.php?topic=5432073.msg61507211#msg61507211): it has a lot of "generic text", and some parts are just bad practice and shouldn't be advised. All in all, it's pointless. But it's well-written and not against the forum rules.

Same here. I have also started to suspect the same. I have been monitoring one account for the last few weeks, and within this time, the account have posted 3/4 times only which seems like AI generated to me. I don't want to publicly post that until I'm sure that he is posting AI generated text.
This will be somewhat a disaster if someone with a long list of account starts to participate in signature campaign and post AI generated garbage, I'm pretty much sure we will see the day in the coming future.

Even if the text-creating-AIs become paid services, it's probably not going to be more than a fraction of a dollar per text, and since many campaigns pay up to a few dollars per post, that's very profitable.
Automating the whole thing will give you passive income. If enough people do this (of if one person creates enough accounts), it will outcompete humans from signature campaigns.

Then this will happen:
Maybe Bitcointalk.human will work as a domain. With a captcha :P

Or maybe the internet isn't meant to last for humans after all:

In the specific case I mentioned, the OP admitted that the post was written exclusively with AI, and after the report, that thread was locked. Of course, it is difficult to prove that someone is using AI for their posts, but I think that the post history of each member can reveal any significant change.

What also interests me about these AIs is how creative they are in the sense of whether they answer the same question 100% identically or still make a difference. It would be interesting if we could find some unique pattern that would indicate that behind such posts is not a human but an AI.

I played around for a while asking Chat GPT whether this or that content was generator by it, and there are some certain circumstances where it declares having come up with the text, or provides the "original text" (or a variation) if the data you feed it differs somewhat. The post you referenced (https://bitcointalk.org/index.php?topic=5432035.msg61505927#msg61505927) was verbatim, so it was an easy match for it to find, but the more you twist it the sooner it declares it not being his. Obviously this is only an indicator though, not a viable declaration of guilt that can be accepted as proof.

If you ask it the same question multiple times it provides different answers, but the essence, and even some fragments, are the same from one iteration to another. A unique pattern will be hard to find, though it does tend to be rather prosaic, often tries to play the neutral role, and for now will generally lack the human touch to its answers, being rather consistent in its posts, as opposed to what one would really be creating by writing posts with a wide range of tones, content length and style.

My point is that that post is no worse than many of the ones we see out there in terms of low value or spammy, and that the fact that it is written by an AI should not be enough to delete it, considering that there is nothing about it in the unofficcial rules and even more thinking that the OP has made it clear that it is written by an AI.


The one we are having here has nothing to do with gambling, the one over there does, because they are talking about supposed ways of winning in gambling.


Can I know the reason? (low value, spam, etc). Btw I see it's been locked.


I have seen many such posts on other sites and they have in common that they are rather generic and too formally structured, like, point 1, point 2, summary, conclusion, something like that.

My worry is the sheer volume of posts an AI can produce.

Damn it... well spotted LoyceV ... I asked the ChatGP similar question and look at the answers


It looks all the four bullet points are somehow inspired by ChatGPT

First bullet point coming from Tip No. 5 of ChatGPT rephrased
Second bullet point coming from Tip No. 2 of ChatGPT copy paste + rephrased
Third bullet point coming from Tip No. 3 of ChatGPT copy paste + rephrased
Fourth bullet point coming from Tip No. 4 of ChatGPT rephrased


https://i.imgur.com/mLlGnFc.png

Give that man a cookie for using ChatGPT :)

Playing devil's advocate: it's still possible the ai chat learned from the recent post. I think it's unlikely, but the possibility can't be rules out. Damn ai!

Ah yes, I suppose I'm just a relic of the past. But I still believe that an internet forum should be a place where humans come together to discuss and share their thoughts, not a robot convention where mindless machines spout pre-programmed responses. Real discussions require actual thought and consideration, something that robots just don't have the capacity for. 

I don't think we need any specific rules about what constitutes an "Internet forum," because it's one of the most basic definitions of that term:

I actually messed around with ChatGPT a few days ago, and I find it both outstanding and scary at the same time. It can generate anything, from articles to code, while finding answers to your most stupid questions. On the one hand, it can be an extremely useful tool to use in projects and assignments, for instance, in moderation of course. On the other hand, it can easily be abused, from posting in the forum to even using it to generate massive walls of text in a master's thesis. AI is getting out of hand.

Good job, LoyceV, for spotting AI posters. Despite having seen how the application works, it didn't cross my mind that it was AI-generated, and truth be told, I've seen quite a few similar ones the past few days but thought nothing of it.

So maybe I just used my own reply for bullet points of my own thread. Or I asked chatgpt for what reply I should post on this thread - https://bitcointalk.org/index.php?topic=5431962.msg61503560#msg61503560 (https://bitcointalk.org/index.php?topic=5431962.msg61503560#msg61503560)

I learned about chatGPT from a youtube video published by a YouTuber Gotch SEO and later that day on this forum. I asked chatGPT many questions but not this one.
Besides, I see only one point is matched which is also I replied to someone's post already without bullet point before creating my own topic.


I want that cookie for sure.

There is a close resemblance indeed, though I couldn’t get Chat GPT to declare having written a single paragraph, nor selected line within. It could have been span and custom tailored to deviate from the original suggested output, but then again, it does come up with some similar conceptual answers when denying being the author:

The proposed strategies are though, as expected, more or less common knowledge, and there could be different sources that agglutinate them and that were used as an inspiration. Whether the inspiration was for the poster or Chat GPT is what remains unclear.


Allegedly, it can't use recent data. According to "it":
I figure though that Chat GPT is based on the GPT-3 model text-davinci-003, whose training data ran up to Jun 2021 (as per OpenAI’s model overview (https://beta.openai.com/docs/models/overview)).

If you mean what I stated in the report, then it is "low value/zero effort post generated by AI". Of course it was locked, and quite quickly if I may say so - I don't think anyone wants to encourage forum members to use such tools to publish posts.

Bear in mind that ChatGPT has a limited general use when directing through prompts.

Given a non-zero-shot set of prompts, catered to BCT, you could very easily create an engine for post generation. Though, it is magnitudes more expensive than GPT-3 Davinci (if you're paying). If you had enough resources, though, you could easily sidestep the OpenAI cluster and try to use other models (e.g. EleutherAI (https://www.eleuther.ai)).

The problem is that how will you proof that a message is AI written or self written ?

One way to go about is that you will find all of his posts in a similar pattern of AI generated text but what if he uses AI in some of his posts and he writes other posts himself. Then i guess it will be even more difficult to judge when there will be mixtures of AI and self written text.

The value of AI generated posts actually depends on what it's fed with, if you feed it garbage it generates garbage (garbage in garbage out).
I've gone through several posts generated by it and I think they are mostly worthless, inconsistent, full of lies etc, to the extent I asked it who is his creator and it said a car marker & others created it... "No wonder you are not so smart" I told the AI. I was embarrassed and stop using. Will probably start looking for an alternative or we could join hands together to create a far more superior one based on truth in less than a month.


From my observations of existing AI on the internet, I suspect they look for very honest people or sites in existence and stick to them for Holy/Pure information in order to consistently generate good information that's very dependable and helpful to society without issues. I hope they will be beneficial to the sources rather than being parasitic in the manner they feed from them.

There are tools to detect this. You can find them by simple Google search.
But the problem will be of scale. If people start abusing this here or on any other forum, because of the huge volume of spam, it will consume lot of forum staff time to spot and end this. This challenge will be faced by all forum owners in future. One site I can give example of here is Quora which is getting filled with such AI posts.

There are, but at least the ones I’ve tried out ,that allow you to perform a free trial online, have seemingly very poor results detecting GPT-3 based Chat GPT text, and that’s just feeding a direct output as input to the detector, with no spinning or customization at all in between to even try to throw them off their tracks. There are a few additional ones I would have been prone to try, but they required subscription and payment (a no go at this stage).

Were the outputs to eventually have an inbuilt digital watermark (https://bitcointalk.org/index.php?topic=5432184.msg61513651#msg61513651), there could be an easier basis to detect AI generated type text, but even so, each company will have used their own method to generate the (potential) watermark, and I’m pretty sure that adding a step or two in the end-user’s pipeline would largely degrade the watermark’s persistence and utility (though that requires some minimal effort on the user’s side, which is not a done deal).

Indeed, not all services are reliable and free at the same time. Even if that were the case, we simply couldn't go around to each forum post and test if it was written by AI. There are some hints that may indicate that a post was written by AI, but my main concern is that services such as ChatGPT may even create duplicate content if the same parameters are used, and it may cause post bursts and spam issues since it is capable of producing a wide variety of posts within a short period of time. AI may not produce low-quality posts, but it'll definitely create non-engaging discussions, which in some cases, may be worse than plagiarism.

Keep in mind that ChatGPT will never use the colloquial "we" when writing an answer (they might say I) unless you explicitly tell it to. So posts with mostly AI sentences but a few human-generated sentences are like a semi-automatic gun: it's not a full-fledged machine gun, but still dangerous nonetheless, and more mobile.

That intrigued me to try this "we" thing and it worked :D


https://i.imgur.com/6WeTbws.png

See? Now it's easy to tell a post written by a robot if not by a complete moron.

What does Chatgpt thinks of the future of Bitcointalk in cryptocurrency?
   https://i.imgur.com/VbSmUFo.jpg          https://i.imgur.com/kY7suXf.jpg


he predicts more competing platforms will appear in the future, making the platform less relevant

Phew... I'm safe coz GPT-3 isn't available in my country although I could probably do it with VPN, I'm too lazy to do it anyways. Not to mention I feel bad that my writing isn't genuinely coming from me.  Good thing that we can use tools to detect and hopefully we got enough manpower to be able to catch all of them.

ChatGPT won't be free forever, they're spending more than 100k usd per day on API calls
at some point they'll probably start to charge for it which will change the dynamics a bit

but still important to discuss it here since the price to use it will probably still be lower than what signature campaigns pay here in the forum so people would still have an incentive for arbitrage. (creating low effort posts using AI)

There are countermeasure now coming and the first one which I saw is GPTZero, an app that detects AI-written essays. Many students are using ChatGPT for doing there college assignments and this is negatively impacting there ability to think on there own. Once you pass your school/college then one can use such tools in professional life.
 
https://i.imgur.com/UOYeD5n.png
Image source (https://www.polygon.com/23540714/chatgpt-plagiarism-app-gptzero-artifical-intelligence-ai)

Chat GPT3 is only the beginning: OpenAI is working hard to improve their product:

From LinkedIn (not providing link because of opec).

https://i.ibb.co/GQMH1r1/61581946.jpg

I'm pretty sure GPT4 will not be free and will be available only through a subscription.

Meanwhile I am still waiting for OpenAI to get back to me regarding my spot on the ChatGPT professional waitlist.

I saw on twitter a lot of examples where the AI can produce text that are perfectly detected as human. This is done is several ways: adding smileys after every word, adding slight typos or double commas.
Of course it’s a mice and rat race in these kind of techniques.

wow, GPT4 will be crazy to see

for those interested in checking more information about GPTZero verification tool there's a good viral thread on twitter
here:
https://twitter.com/edward_the6/status/1610067688449007618?s=27&t=dP7KaJjQT-vSn1y2jz_32A

Which will limit the problems that everyone's having with the technology. The thing is, something like this when used appropriately can help in several fields. It's the abuse that comes with that, which is unfortunately taking away from the limelight of genuine usages. For example, it's pretty incredible to find documentation on certain snippets of code, much faster than using traditional search engines.

The subscription will be one roadblock to misuse, but also the tools coming out to verify it. ChatGPT has already talked about implemented some sort of identifiable system, which is largely the case now it just needs to be fine tuned, and verifiable via their system rather than third parties. Honestly, I don't think we'll have a huge problem going forward once these things come into place.

+1.

If this is a forum where legitimate users are supposed to express their opinion, why should we be reading what someone has requested from an AI and then copied and pasted here?

What I like about Bitcointalk is reading people's opinions. Not what the AI returns to them in order to increase their post count, merit or whatever. Don't be fooled, you already know the main reason why someone would consider doing it...

An interesting read from the meta section of stackoverflow where they managed the same problem and is full of opinions on this topic: https://meta.stackoverflow.com/questions/421831/temporary-policy-chatgpt-is-banned

Sorry to re bump this thread but i see today some people... how can i say they dont think or for them doesnt matter the health of the forum.

I say this because in one post of the beginers sections, i clearly detected a IA account and im not an specialist in this, but the sad part its after some more usser found and claim this (and really wise users) some other keeps coming and talking with this IA user.... come on man at least read the thread.

With GPT4 i dont know what to expect because can be a lot of hype or this can go beyond us imagination.

We will soon discover it as GatGPT4 is expected to be launched next week, apparently:



GPT-4 release date and more: Everything we know (https://www.ghacks.net/2023/03/13/gpt-4-release-date-and-more-everything-we-know/)

I guess that even if the improvement will be outside the language models, the interpretation of input will be orders of magnitude better than j it currently is.
We will soon be replaced by _{yet another} bitcoin omniscient AI.

BPIP should include an feature to identify IA's. Maybe modify the extension to put a badge under their name that says "Suspected bot account".


Yeah but it will take months before the services that make use of GPT4 are generally available to the public, including those form OpenAI itself.

The new ChatGPT-4 was released yesterday and is available through the GPT-4 Plus paid plan. ChatGPT is said to have been improved and can now also present images to the text. Nevertheless, the improvements are probably not as good as hoped (which I think is a good thing). Either way, the changes from the previous ChatGPT 3.5 are very minor, according to Sam Altman, the CEO of OpenAI.  :)

https://i.imgur.com/JJBKytJ.png
Source: https://twitter.com/sama/status/1635687853324902401

As long as the CEO is human, I'm not impressed.

Sam Altman will keep bringing in newer versions within shorter intervals because information is never static and i doubt if there will ever be a flawless openAI. Deep learning/transfer learning is yet to out grow the phase of negative transfer learning on the models been feed on. A mirror change is what we keep getting

If there is any industry the ChatGPT-4 will long struggle with is the blockchain because protocols keep changing as well as newer invetories and use case. Building models on everyday experience can have a long time positive impact but on a consistently improving network it might be a lifetime of discoveries

I am sure he's trying to downplay the relevance of this upgrade.

There are people already building businesses (https://twitter.com/linusekenstam/status/1635754587775967233?s=46&t=ybCp3ydjDJA_wR_uVxrEgA) on this new version.

:D :D :D  Hopefully it stays that way.  :D

---

That's quite possible. I haven't been all that interested in this topic until now, but since it got here in the forum I've been keeping an eye on it.
I'm curious when the next version will be released and how long it will take.  :(

---

That could be. Since I don't use ChatGPT, it's hard for me to rate it.

Thank you for your hint. That was really quick.  :o