Pages:
Author

Topic: Bounty spammer meets AI Chat generator. Can they be banned on the spot? - page 4. (Read 2557 times)

sr. member
Activity: 756
Merit: 390
But these systems have a lower quality level than OpenAI, making it easier to detect text created by AI. Now, of course, it will not eliminate this situation, but the fact that this system is paid now, will at least reduce the most difficult cases to detect.

The free subscription is still available found this on their FAQ.

Quote
Yes, free access to ChatGPT will still be provided. By offering this subscription pricing, we will be able to help support free access availability to as many people as possible. See our general ChatGPT article for more information on our free offering.

Link to FAQ


I can still access my dashboard and start a new chat. You can check the screenshot.



The main reason behind the subscription plan can be found in FAQ. I am quoting the explanation below.

Quote
ChatGPT Plus is a pilot subscription plan for ChatGPT. It offers availability even when demand is high, faster response speed, and priority access to new features.

I agree the other AI content writers that are free are not as good as OPEN AI but if you know the trick then you can create content just like a human writer. I have seen many papers submitted to me and I would not say that it is easy to detect AI-based content but it takes a lot of time an effort to find one. Still, you cannot guarantee that the content was written using an AI. 
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**
Apart from OPEN AI, there are other projects that offer free AI on a trial basis. I do not wish to reveal their names but the trick is to use them for free is just register using a different email address whenever the free period expires. These content writers are also very good and effective. If you think this behavior would get reduced then you are wrong. I say this because of my experience in the teaching profession I know how and through what my students have been submitting assignments.


But these systems have a lower quality level than OpenAI, making it easier to detect text created by AI. Now, of course, it will not eliminate this situation, but the fact that this system is paid now, will at least reduce the most difficult cases to detect.
sr. member
Activity: 756
Merit: 390
OpenAI is already being paid for! It should now reduce a little this type of behavior here in the forum.

Either way, the price isn't too high: $0.0004 / 1K tokens (low cost); $0.0200 / 1K tokens (high cost).

Quote
Prices are per 1,000 tokens. You can think of tokens as pieces of words, where 1,000 tokens is about 750 words. This paragraph is 35 tokens.

https://openai.com/api/pricing/


Apart from OPEN AI, there are other projects that offer free AI on a trial basis. I do not wish to reveal their names but the trick is to use them for free is just register using a different email address whenever the free period expires. These content writers are also very good and effective. If you think this behavior would get reduced then you are wrong. I say this because of my experience in the teaching profession I know how and through what my students have been submitting assignments.
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**
OpenAI is already being paid for! It should now reduce a little this type of behavior here in the forum.

Either way, the price isn't too high: $0.0004 / 1K tokens (low cost); $0.0200 / 1K tokens (high cost).

Quote
Prices are per 1,000 tokens. You can think of tokens as pieces of words, where 1,000 tokens is about 750 words. This paragraph is 35 tokens.

https://openai.com/api/pricing/
legendary
Activity: 2072
Merit: 4265
✿♥‿♥✿
Yesterday I came across an article with a link to a site that is supposedly a classifier for identifying texts written by AI tools.
But as the article says, it's still not an accurate tool. In the testing process, texts written by AI were identified by only 26%, and also erroneous by 9% in texts that were written by a person.

https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text/
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**
Current predictive models more or less generate texts word for word. Simplistically put, at each point there's a list of weighted candidate words, one of which is stochastically chosen (according to some distribution). The weights can be gently adjusted with a "fingerprint", so that the algorithm balances similar words with a statistically detectable frequency (the "fingerprint"). In texts with more than a thousand words, the probability that this would occur randomly is low (depending on how strongly the "fingerprint" was allowed to alter the weights).

GPT-2 has a loophole! You only need a spin writer on the text that has been generated by Chat GPT or any other AI content bot and it will not be able to detect. You might be only finding those members who do not have any knowledge of the English language. Those lazy buggers who know the language but do not want to write on a topic that they think can be posted here cannot be tracked. I would not rely on a free tool to check AI-generated content. There are some paid ones but it is not worth investing in as long as the content makes sense and does not break any rules of the forum.

Don't forget that several AI systems write in several languages. So, the response has to be very mechanized/generic, presenting information out of context so that we can understand whether it originates from AI or not.

I agree that investing in research tools is very risky and not viable.

I think the best way to find out if the post is made using AI is by asking the user about their response. Ask who explains the same. The way it responds (if it responds) can hint at whether or not an AI is responding.
sr. member
Activity: 756
Merit: 390
How do you think we are detecting plagiarized text by other members?
We certainly don't have super powers, and I was always using two website to detect source and PERCENTAGE of plagiarism... it worked well most of the times.

EDIT:
Here is one test I did with this post from member Techdev generated by AI (it was later deleted by moderators after my report):
https://ninjastic.space/post/61619530

GPT-2 Output Detector is showing 99.98% that this is AI generated:



GPTZero is saying that text is most likely to be AI generated:



GPT-2 has a loophole! You only need a spin writer on the text that has been generated by Chat GPT or any other AI content bot and it will not be able to detect. You might be only finding those members who do not have any knowledge of the English language. Those lazy buggers who know the language but do not want to write on a topic that they think can be posted here cannot be tracked. I would not rely on a free tool to check AI-generated content. There are some paid ones but it is not worth investing in as long as the content makes sense and does not break any rules of the forum.
copper member
Activity: 42
Merit: 67
Current predictive models more or less generate texts word for word. Simplistically put, at each point there's a list of weighted candidate words, one of which is stochastically chosen (according to some distribution). The weights can be gently adjusted with a "fingerprint", so that the algorithm balances similar words with a statistically detectable frequency (the "fingerprint"). In texts with more than a thousand words, the probability that this would occur randomly is low (depending on how strongly the "fingerprint" was allowed to alter the weights).
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**
Perhaps when there is a difference between the paid and free versions, this supposed watermark will appear in the free versions.
So spammers can pay to remove the plagiarism marker? Sounds like a great business model!

I first question whether the texts created by these AI really have any "watermark".
If they really have, I wouldn't be surprised that the paid versions of these AI would no longer include that watermark.

The only way for a text to have a watermark is a word, phrase or other content that is the same in all texts, regardless of the subject being treated. Honestly, I have a hard time seeing any other way this could happen, but I'm not an AI expert, so it might be different.
hero member
Activity: 1659
Merit: 687
LoyceV on the road. Or couch.
Perhaps when there is a difference between the paid and free versions, this supposed watermark will appear in the free versions.
So spammers can pay to remove the plagiarism marker? Sounds like a great business model!
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**
Few AI has invisible watermark[1] which can be used to identify whether a post is written with AI or not, although i doubt moderator would bother use tool to detect such invisible watermark.

First we needed to know what that watermark was. And honestly in text, I always find it a little difficult to use.

Unless in the middle of the text standard information is inserted so that it is detectable in a copy-paste scenario. But the first step has to be for the company that develops that AI to introduce that information. Perhaps when there is a difference between the paid and free versions, this supposed watermark will appear in the free versions.
legendary
Activity: 2716
Merit: 1225
Once a man, twice a child!
Incidentally, that's the price to pay for such tech advancement. Anything with advantages has its demerits too. Recently, off this forum though, I've been reading how well a certain Chatgp Al has been performing and dishing great stuff. Humanity, at some point, will have a serious competition with computer intelligence. Nonetheless, I trust theymos will come up with an idea that can put all this Al activities on the forum at bar. We need real humans to chat and post here, not some Al. By the way, nothing judges a situation better than human analysis. Look at those sampled posts from alleged Al and see how off point they're because of the lack of emotional perspective to those comments.
newbie
Activity: 12
Merit: 0
I was just experimenting with by whom I will be caught first. LoyceV, you're quick.

Chat isn't strong enough yet. But it's a matter of time.
member
Activity: 119
Merit: 38
Yo! Member
Running a pruned node has downsides for users.
~
In the event of a dispute, it can be more difficult to prove the validity of your transactions.
I'm thinking about neutral tags for chat bot BS.
But it will be difficult to conclude. These chat bots are as smart as the forum AI  Tongue
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**
My question to everyone who advocated banning them is, how would you know that the reply is being done using an AI or a topic is created using the help of AI?

I think it is almost impossible, if the use of this tool is well done.
If it's a matter of just copy-pasting, then what might happen is when we figure out how the question was asked the result is the same.

We must not forget that there are users who can use translation tools, and that sometimes builds the sentence in a less "natural" way. I for example write using a translator.   Lips sealed
hero member
Activity: 2156
Merit: 803
Top Crypto Casino
I think a moron posted those replies! If he/she knew how it worked then it would not have been noticed.

Yet, it is a pretty serious issue and I solution has to be found. I have seen many topics posted on many boards which feel that they are being posted using an AI. As most of them do not make any sense. I have reported them to moderators and still those replies stay.

I have done some research to find a solution to those AI replies but they are not conclusive. Any steps like banning might not be inconclusive as you can not prove that someone has used an AI.

My question to everyone who advocated banning them is, how would you know that the reply is being done using an AI or a topic is created using the help of AI?
hero member
Activity: 1659
Merit: 687
LoyceV on the road. Or couch.
Running a pruned node has downsides for users.
~
In the event of a dispute, it can be more difficult to prove the validity of your transactions.
I'm thinking about neutral tags for chat bot BS.
legendary
Activity: 1722
Merit: 4711
**In BTC since 2013**

I did some simulations, to try to understand if what is said is similar to the AI.
Of course, I don't know how the AI research was done, even so some posts seem identical to your answers or the way to respond. But on the other hand, other posts don't seem like that anymore.

We had to remain attentive and watch the evolution of that user's posts in order to obtain greater clarity on the subject.
hero member
Activity: 1659
Merit: 687
LoyceV on the road. Or couch.
legendary
Activity: 2436
Merit: 1189
Need Campaign Manager?PM on telegram @sujonali1819
After launching the ChatGpt it's a benefit for the spammer and the beginner who doesn't care about bitcointalk rules. Now I can feel what would if the merit system did not arrive. Theymos did a great job by implementing the merit system. Otherwise, bitcointalk could be full of spammers with the help of this AI chat generator.

Btw I also think the spammers who use the AI chat generator to post here should be at least a temp ban. Otherwise you can not stop them it's sure.
Pages:
Jump to: