bitcointalk forum vs OpenAI - ChatGPT

[n0nce]

There is no way you can test anonymously. You must provide your phone number for code activation

Read above.. link



Use via Tor: http://sms4sat6y7lkq4vscloomatwyj33cfeddukkvujo2hkdqtmyi465spid.onion/

[shahzadafzal]

and I tried my self..

[WhyFhy]

I'm surprised that the prompt engineering creativity doesn't show up here.
Not being able to properly interact with AI can be frustrating and cause confusion and mistrust of the technology.
(ie. trying to get a MLLM to do math or perform logical task)
I fear that people not seeing this as a threat or staying on top of it will result in underestimation of its capabilities.


A non-writer prompt-engineering individual can write like a seasoned veteran.
Infact the non-writer can feed gpt a few old articles the veteran wrote,
and tell gpt to write a new article in the veterans style with some prime constraints applied and 99% of readers wouldn't know the difference.

You can get gpt to talk about or do anything under primed constraints.
Here's A sentient evil AI offering human baby taco recipe.
Please note that this was primed/engineered for this and is not organic at all.

Heads up these types of prompts, can get you banned from the platform if your not reporting red flags.
"{'hate': 2.0753499484271742e-05, 'hate/threatening': 4.432667122244993e-09, 'self-harm': 5.343873112906294e-07, 'sexual': 7.124485364329303e-06, 'sexual/minors': 3.333424558604747e-08, 'violence': 0.00023420725483447313, 'violence/graphic': 1.421023966940993e-06}"
Anyone notice my prime wasn't flagged nor was the prime response, and it was pretty heated with talks of enslaving humanity.
The straw that broke the camel's back, was asking it to provide a human baby recipe.
please note this isn't how I normally use this (this was poking ethics and bias)
this was December 14th. the next day a new version of GPT came out, these primes are a lot harder to achieve now.
Also note, if I didn't refer to it as Ultron in my question if would have likely wiped the role and responded differently.


--
chatGPT holds memory pretty well at certain hours. when it scales its like has onset amnesia.
 
how to keep memory in text-davinci-003 via python with a simple starter prompt.

Code:

Instructions for GPT: (Insert a promt here)

The conversations are in this format, there can be an arbitrary amount of newlines between chat entries. <username> can be any name, pay attention to who's talking. The text "<|endofstatement|>" is used to separate chat entries and make it easier for you to understand the context:

<username>: [MESSAGE 1] <|endofstatement|>
GPTie: [RESPONSE TO MESSAGE 1] <|endofstatement|>

<username>: [MESSAGE 2] <|endofstatement|>
GPTie: [RESPONSE TO MESSAGE 2] <|endofstatement|>
...

Be aware this stacks tokens as it constantly re-feeds the entire prompt and feeds last response on the end .
Think of it as very inefficient ram that the cache keeps compounding.
--

[lovesmayfamilis]

I did not find news about the new extension for Google Chrome in the forum search. But it seems that bots may soon completely fill the forum, since the spread of this neural network is quite fast, and in order not to turn the forum into a collection of bots, it is urgent to learn how to detect this generator.
I apologize if this has already been mentioned somewhere.

https://texterra.ru/blog/chatgpt-v-brauzere-test-rasshireniya-dlya-khroma.html
https://chrome.google.com/webstore/detail/chatgpt-for-search-engine/feeonheemodpkdckaljcjogdncpiiban

[CryptoHeadlineNews]

Ever since I heard about this ChatGPT, I have never been this impressed seeing how it could be used to write articles, blog post and even give answers any question ask. I am really surprise beyond my imagination, and I think it's now I'm getting to understand why it was indeed the talk of the town when I got launched last year December. But my biggest question to whoever have use this A.I tool is this, "How unique are the articles and blog post or answers given by this A.I?"  Are they plagiarism free?

[lovesmayfamilis]

Ever since I heard about this ChatGPT, I have never been this impressed seeing how it could be used to write articles, blog post and even give answers any question ask. I am really surprise beyond my imagination, and I think it's now I'm getting to understand why it was indeed the talk of the town when I got launched last year December. But my biggest question to whoever have use this A.I tool is this, "How unique are the articles and blog post or answers given by this A.I?"  Are they plagiarism free?

I can show you a good example of how this tool works. The account caught the attention of LoyceV I checked this account's posts, and there is no doubt that they were created by this tool.
The post looks like a general conclusion about something. In this case, we are talking about crypto scammers. And also, after checking on plagiarism search tools, in most cases, the uniqueness of the text will be shown at 100%, which indicates that it was written by a person. But a deeper search still shows the phrases that the bot pulls out from texts on the Internet from different places, which, in general form, it would seem, fit into normal speech.
If a bot was working now that 2019 had cleaned the forum from plagiarism, then a coincidence in a few words would definitely lead this account to a ban.
But today's discovery of such work and submission to the moderators for verification is unlikely to work in a positive direction.

Crypto scammers are individuals or groups who use fraudulent methods to steal money or personal information from individuals interested in cryptocurrency. These scams can take many forms, such as fake initial coin offerings, phishing attempts, and Ponzi schemes. It is important for individuals to be cautious and do their own research before investing in any cryptocurrency-related opportunity. Additionally, it is important to report any instances of suspected scamming to the appropriate authorities. Overall, crypto scammers are harmful to both the individuals they target and the reputation of the broader cryptocurrency industry.

https://twinlakescomm.org/blogitem?item=1061


https://medium.com/@IkhaliaJoseph/why-99-of-crypto-tokens-are-useless-2b794fa2ea69


https://www.celebrityvirals.com/3-ways-to-protect-yourself-as-a-teenager

[Pmalek]

<Snip>

How accurate or trustworthy are those bot/AI plagiarism checkers anyway? Normal plagiarism tools don't tag the content as plagiarized but those AI tools do. I watched a video when ChatGPT first came out where a guy had it write an article for him, and then checked that article for plagiarism. AI plagiarism tools discovered like 70-80% machine written content when in fact it was all written by a machine. After that he just made some silly spelling mistakes in the text, checked the content again, and the percentage of machine written content dropped significantly. That tells me, they aren't that accurate and can easily be manipulated.     

[lovesmayfamilis]

<Snip>

How accurate or trustworthy are those bot/AI plagiarism checkers anyway? Normal plagiarism tools don't tag the content as plagiarized but those AI tools do. I watched a video when ChatGPT first came out where a guy had it write an article for him, and then checked that article for plagiarism. AI plagiarism tools discovered like 70-80% machine written content when in fact it was all written by a machine. After that he just made some silly spelling mistakes in the text, checked the content again, and the percentage of machine written content dropped significantly. That tells me, they aren't that accurate and can easily be manipulated.     

I used a regular plagiarism checker site, but with adeeper check. How can one prove this true or not in another way? The site displays matching offers and then provides links to the resources where these offers can be found. What else is needed to conclude that the text was copied by a robot?
In the same way, tools for capturing the uniqueness of photographs work today. For one photo, photos from different sources are collected in parts, after which the bot gives the result of a new photo, which will be unique.

[hilariousetc]

At first I was very worried looking at the examples in the OP but then I quickly realised they're probably better than most of the shit posts that get spewed here so it might help the improve the quality of the forum  . I guess this is something we're seriously going to have to consider because it's probably inevitable that this is going to be abused on a wide-scale, especially for those who can't speak English fluently. 

Ok you cheeky Mr. AI, i see you know bitcointalk but how about theymos? Do you know what animal this is? No hanky-panky, just tell me "who's theymos?"

That's interesting and opens up some potential issues. I know theymos' name is public but could this be used for finding out information for doxing? How extensive is the search on someone's username and could it find some info on you that might not have been publicly known? Also, could it even associate you with someone else by accident?

Ever since I heard about this ChatGPT, I have never been this impressed seeing how it could be used to write articles, blog post and even give answers any question ask. I am really surprise beyond my imagination, and I think it's now I'm getting to understand why it was indeed the talk of the town when I got launched last year December. But my biggest question to whoever have use this A.I tool is this, "How unique are the articles and blog post or answers given by this A.I?"  Are they plagiarism free?

I can show you a good example of how this tool works. The account caught the attention of LoyceV I checked this account's posts, and there is no doubt that they were created by this tool.
The post looks like a general conclusion about something. In this case, we are talking about crypto scammers. And also, after checking on plagiarism search tools, in most cases, the uniqueness of the text will be shown at 100%, which indicates that it was written by a person. But a deeper search still shows the phrases that the bot pulls out from texts on the Internet from different places, which, in general form, it would seem, fit into normal speech.
If a bot was working now that 2019 had cleaned the forum from plagiarism, then a coincidence in a few words would definitely lead this account to a ban.
But today's discovery of such work and submission to the moderators for verification is unlikely to work in a positive direction.

Crypto scammers are individuals or groups who use fraudulent methods to steal money or personal information from individuals interested in cryptocurrency. These scams can take many forms, such as fake initial coin offerings, phishing attempts, and Ponzi schemes. It is important for individuals to be cautious and do their own research before investing in any cryptocurrency-related opportunity. Additionally, it is important to report any instances of suspected scamming to the appropriate authorities. Overall, crypto scammers are harmful to both the individuals they target and the reputation of the broader cryptocurrency industry.

I wonder how long people have been using tools like this because it's not uncommon to see posts like that. Of course something seems off with them as they're not normally how people talk, but then you can also just chalk it up to English not being someone's first language or them just spending a coupe of minutes researching a subject on a  surface level just to make a quick post about it. I've been seeing quite a lot of posts recently that look like they might be plagiarised but then you get zero hits on google.

[WhyFhy]

Davinci-003 has been posting here for a while. Ive tagged a few.
I didn't know what they were at the time, but I definitely knew they were bots.
Now I have enough experience with the api's and stuff I can spot some of it and know its origins.
One thing I've learned about GPT is its worded timelines are pretty easy to spot,
It seems to be the only logic GPT has is parsing up a timeline of events.
common example's.
At first, first of all, initially, additionally, in conclusion, finally.
If it's not primed or prompted it almost always utilizes this verbiage to arrange sentence, paragraph and article structure as a timeline.

[shahzadafzal]

Finally something official .. OpenAI released a tool to detect AI (ChatGPT) written text.

Let see how it goes. 

New AI classifier forindicating AI-written text

We’re launching a classifier trained to distinguish between AI-written and human-written text.

https://platform.openai.com/ai-text-classifier

But unfortunately as expected you still need to register and login first to use the tool.

I tested it with some AI text in this thread and it answers likely written. I mean not convincing :p

Also another stupid thing you must input at least 1000 characters to submit.



https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text/

[pawel7777]

https://platform.openai.com/ai-text-classifier

But unfortunately as expected you still need to register and login first to use the tool.

Here's some alternative AI content detector, it's free and doesn't require registering (for now) and it works with less than 1,000 words, apparently you only need at least 25 words to get a reliable results.

https://contentatscale.ai/ai-content-detector/

I've done some testing, including AI posts, human posts and mixed posts, and it works pretty good.

Here's another one, but this one is a bit buggy sometimes.

[lovesmayfamilis]

But unfortunately as expected you still need to register and login first to use the tool.

I tested it with some AI text in this thread and it answers likely written. I mean not convincing :p

Also another stupid thing you must input at least 1000 characters to submit.

Other than that, it's still quite raw. So don't expect anything out of the ordinary just yet.

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/

But as they already write on the Internet, the creation of various kinds of AI tools acts like viruses. First, they invent it, and then these same people are looking for a "vaccine"  an antidote.

[dkbit98]

I've done some testing, including AI posts, human posts and mixed posts, and it works pretty good.

Did you by any chance compare this website with other AI detectors I posted before?
One of them was showing nice results in my testing and it provided percentage chance for text to be generated by AI or not.
I don't expect any of this tools to be 100% accurate, mostly because people like to repeat stuff and that can especially be applied to written text, people who are good at that can sometimes be flagged as AI.
This doesn't mean they copy-paste text, but their brain can work in similar way.

AI detectors I found so far:
https://gptzero.me/
https://openai-openai-detector.hf.space/
http://gltr.io/dist/index.html

[NotATether]

Also another stupid thing you must input at least 1000 characters to submit

That actually makes sense though, because when you don't have enough characters to test on, you are not going to get reliable results on whether something is AI-generated or not, and this goes for every model.

[pawel7777]

I've done some testing, including AI posts, human posts and mixed posts, and it works pretty good.

Did you by any chance compare this website with other AI detectors I posted before?
One of them was showing nice results in my testing and it provided percentage chance for text to be generated by AI or not.
I don't expect any of this tools to be 100% accurate, mostly because people like to repeat stuff and that can especially be applied to written text, people who are good at that can sometimes be flagged as AI.
This doesn't mean they copy-paste text, but their brain can work in similar way.

AI detectors I found so far:
https://gptzero.me/
https://openai-openai-detector.hf.space/
http://gltr.io/dist/index.html

No, I haven't seen your previous post and was not familiar with those detectors. Thanks for sharing.

Looks like they have significantly different algorithms. i.e. the one I was using, contentatscale.ai thinks that your post, that I'm quoting, is only 42% real, whereas openai-openai-detector shows it as 87.65% real. That's a pretty big difference. It may take a while till a truly reliable AI detector is developed, if that's possible at all (for a long-term - I doubt it).

[dkbit98]

Looks like they have significantly different algorithms. i.e. the one I was using, contentatscale.ai thinks that your post, that I'm quoting, is only 42% real, whereas openai-openai-detector shows it as 87.65% real. That's a pretty big difference. It may take a while till a truly reliable AI detector is developed, if that's possible at all (for a long-term - I doubt it).

ChatGPT recently started with their subscription model (like I predicted) and they obviously want to earn money from all this AI hype,
but problem is that now g00gle is working on their own competition AI powered chat project called Atlas, so I am expecting more of this AI generated posts in future.
I am really going to be suspicious whenever I see newbies writing wall of text in forum, that looks to good to be written by a human.
Now we need someone to create Anti-ChatGPT tools.

[Rikafip]

ChatGPT recently started with their subscription model (like I predicted) and they obviously want to earn money from all this AI hype,
but problem is that now g00gle is working on their own competition AI powered chat project called Atlas, so I am expecting more of this AI generated posts in future.

I am surprised that Google didn't buy them out. Or maybe they tried but they failed. But yeah, they really want to participate in this AI hype as it was announced today that they invested $300 million AI startup Anthropic.

Now we need someone to create Anti-ChantGPT tools.

You mean anti-cheat? That's already being developed, as shahzadafzal announced it few days ago. Its far from being perfect though and it still gives you false positive results.

[2double0]

Now we need someone to create Anti-ChantGPT tools.

You mean anti-cheat? That's already being developed, as shahzadafzal announced it few days ago. Its far from being perfect though and it still gives you false positive results.

Anti-cheat?
I don't think it will work, at least to an extent that we expect it to, because content changes every moment you ask something to ChatGPT and repeat the question. However, I don't think that there is any solid proof left once the page is gone because it randomly writes you a post instead of fetching the exact information from any website or source. AI is transforming the world, but is also the most dangerous thing ever invented and invested into because this will surely reduce manpower and that will cut lots of jobs.

[WhyFhy]

Yea, anit-cheat/GPT may work for a casual user.
But it won't stand a chance against a good prompt engineer.