bitcointalk forum vs OpenAI - ChatGPT

[n0nce]

~

Now we're delving into pointless-definition-arguing territory. Let's keep it on-topic: the bottom-line is that AI-generated replies are very often plain wrong; if they are correct, they're pretty generic (so they don't really help in the discussion) and they just sound all the same. They are quite unnatural and unpleasant to read (often also too long).
Since we do have experts on almost any topic, I'd just much rather have an opinion by someone like that who spent years on the subject (even though jotting down the reply takes them just a few seconds - you could count their experience as 'effort'). Instead of getting a semi-correct, not on-point (a bit vague) reply based on an outdated Google snapshot. If you think about it, it's also clear, why: most questions don't have a clear 0 or 1 answer. Therefore, search engines will deliver contradicting results. The AI has no experience in the subject, so it cannot judge which web entries are 'more right' than others; just because one opinion is posted more frequently online, doesn't make it right, for instance. But the AI doesn't know that. Or at least, it cannot judge which answer is true.

Most of the good questions and following discussions on here are about things you can't just quickly web search and find the perfect answer. I think that's the reason why a machine is not well suited for replying in such a forum. If we want 'machine answers', we just search online. The machine simply cannot have a real opinion on a question like I found a paper wallet on a beach ... seriously, because it has no record of something like this in its web snapshot, it has no creative thinking ability or morals. Even though it is called 'AI', it is not actually intelligent.

One example of a 'machine' I really appreciate for very specific 'questions' is Wolfram|Alpha. It is not an 'all-knowing AI', but it just takes natural language input, tells you how it interpreted / structured it and delivers actually correct results based on that structured query. If it cannot do that, it just tells you rather than giving an incorrect answer.

[1714561707]

1714561707

[1714561707]

1714561707

[1714561707]

1714561707

[1714561707]

1714561707

[1714561707]

1714561707

[shahzadafzal]

Most of the good questions and following discussions on here are about things you can't just quickly web search and find the perfect answer. I think that's the reason why a machine is not well suited for replying in such a forum. If we want 'machine answers', we just search online. The machine simply cannot have a real opinion on a question like I found a paper wallet on a beach ... seriously, because it has no record of something like this in its web snapshot, it has no creative thinking ability or morals. Even though it is called 'AI', it is not actually intelligent.

Exactly that is very important point... It's true if a straight forward answer to a simple question is available on the first page of the google then there's no point of post it here unless a user want to increase his post count.

One example of a 'machine' I really appreciate for very specific 'questions' is Wolfram|Alpha. It is not an 'all-knowing AI', but it just takes natural language input, tells you how it interpreted / structured it and delivers actually correct results based on that structured query. If it cannot do that, it just tells you rather than giving an incorrect answer.

Regarding your Wolfram query if you try "Satoshi Nakamoto" it will give an answer but again that's what Wikipedia says

[John Abraham]

https://detectgpt.ericmitchell.ai/

Another AI detector. Pretty good at detecting human text or AI. Maybe more checks are needed, though.

This could be used to earn some money as well. A Campaign manager already offered a bounty for forum users who can catch AI-written text from his campaign participants. I know many members are looking for bounty spammers, cheaters and scammers. I believe they will like this offer and give it a shot. Recently, OpenAI itself released an AI detector (LOL). This is called pure business.

To prevent AI domination effective from today we are introducing an incentive for forum users. Find AI written posts on this campaign
and report to me either in public or in forum PM, please be sure you have enough reference to support the claim. For successful report
the reporter will receive the weekly payment instead of the accused campaigner. The campaigner will be removed immediately.

[bettercrypto]

It helps people get answers for any quarry but will not serve the purpose of google. I learned about it yesterday and was looking for a thread about chatGPT, and you did it already.

My initial thought about chatGPT is like any other AI it collects data from google and delivers the best answer in a way most people like to see. I am going to create an account and will test this later today. I saw many people over-hyped about it, but I don't see any point here. As I mentioned before, it depends on google.

We know that AI chatgpt is a fake robot that acts like a human but has no emotions. It's just like google, the only difference is that it's not updated compared to google.

   That's why when you ask chatgpt about the latest news 2022, he can't answer it properly and properly, because he is no longer updated and the latest news 2022 onwards is no longer in his program, that's what I noticed there.

[shahzadafzal]

We know that AI chatgpt is a fake robot that acts like a human but has no emotions. It's just like google, the only difference is that it's not updated compared to google.

   That's why when you ask chatgpt about the latest news 2022, he can't answer it properly and properly, because he is no longer updated and the latest news 2022 onwards is no longer in his program, that's what I noticed there.

Seriously ChatGPT may be smart but it can be so dumb as you wish it to be... you can guide and mold all it's answer to your wish

Here's a small snippet of the conversation where i presented a situation of a soldier who returned from three year deployment and found wife pregnant but he also had secret a girlfriend 

[Findingnemo]

Maybe little off topic but I wanted to ask about this, can you access their website now cause I wanted to try it and see how it works and decided to bust some AI written post but the website is now and it says we are down and soon will be back so what is happening there is some upgradation going on or just due to lot of traffic the website can't be accessible by everyone?

[nakamura12]

Maybe little off topic but I wanted to ask about this, can you access their website now cause I wanted to try it and see how it works and decided to bust some AI written post but the website is now and it says we are down and soon will be back so what is happening there is some upgradation going on or just due to lot of traffic the website can't be accessible by everyone?

I don't know about your problem but I don't have a problem accessing the website right now as I have posted this reply. Maybe you should try it again and there could be something that the team of that AI is fixing or something like that. I am also curious to know about this AI and I don't think that people should use it as it is not really accurate and there are times that the AI is dumb as what others have stated. Maybe you can access it now.

[cygan]

[RickDeckard]

While browsing the Internet I managed to find a very interesting tool (that probably won't last) : ChatGPT(either version 3.5 or 4) for free[1]! You don't need to link any account of yours, you just go to the website and start chatting! If you want you can find here[2] the source code and some API as well. The owner already received a takedown by OpenAi's legal team so enjoy it while it last. I'm sure many privacy enthusiasts will enjoy this (@o_e_l_e_o and @LoyceV comes to mind).
[1]https://chat.chatbot.sex/chat/
[2]https://github.com/xtekky/gpt4free

[WhyFhy]

Have any of yall tapered off paid subscriptions ,  tried running 3rd party offline LLM's & Stable diffusion models yet?
Models like facebook 30b and openjourney are fun to tinker with.

[Baofeng]

Ok I did try to ask again who is LoyceV as I wanted to see if it has improved from the last time it was queried by the OP,

[img width=450]https://talkimg.com/images/2023/05/16/blob3cc611afa4a47a92.png[/img]

So it's either it has evolved itself, Large Language Models improving by leaps and bounds, their datasets expanding. Or someone from here is feeding it. 

[LoyceV]

Ok I did try to ask again who is LoyceV as I wanted to see if it has improved from the last time it was queried by the OP,

The correct response would have been: "LoyceV is an AI that hates unsophisticated AIs with their verbal diarrhoea".

[Synchronice]

Ok I did try to ask again who is LoyceV as I wanted to see if it has improved from the last time it was queried by the OP,

-snip-

So it's either it has evolved itself, Large Language Models improving by leaps and bounds, their datasets expanding. Or someone from here is feeding it. 

Hmm, this is the response I got about LoyceV today from ChatGPT:


By the way, I asked an AI about me on Bitcointalk and I got this response:

I apologize, but I cannot provide information on a user called "Synchronice" of Bitcointalk.org as there are millions of users registered on the platform, and I don't have access to personal information about individual users. Additionally, Bitcointalk.org is a public forum, and users may choose to remain anonymous or use pseudonyms to protect their privacy.

It's really interesting how this ChatGPT filters things. Btw it thinks that LoyceV is a moderator on this forum and sometimes he is called Loyce. Also, it thinks that Loyce Club helps user to earn merit while it actually does different thing, unites all the tools that Loyce has created for this forum but at the same time ChatGPT understands that LoyceV has created tools and helps the forum.
Sounds interesting and very different from just Google search.

[jvanname]

This is your daily reminder that it is really obnoxious to talk about quantum computing, GPT, AI, and all the other buzzwords without even understanding that reversible computation is the future of computation. And if AI posts become a problem, I suppose we can all sign our messages with our cryptocurrency wallets or talk in a way that AI bots cannot easily mimic. GPT works by stacking a lot of layers. Actually, we don't know how GPT works since it is closed source and because neural networks are not designed to be decipherable. I would trust GPT more if it used more interpretable AI.

[LoyceV]

And if AI posts become a problem, I suppose we can all sign our messages with our cryptocurrency wallets or talk in a way that AI bots cannot easily mimic.

So you're saying we can adjust our way of life to hide from bots. Let's not.

[jvanname]

LoyceV-When the bots become better, we will need to change our behavior. And the main part of this change in behavior is learning what reversible computation is.

[Gladitorcomeback]

Maybe little off topic but I wanted to ask about this, can you access their website now cause I wanted to try it and see how it works and decided to bust some AI written post but the website is now and it says we are down and soon will be back so what is happening there is some upgradation going on or just due to lot of traffic the website can't be accessible by everyone?

Website is working fine but now they have increases the privacy. yesterday my brother tried to open new account and ChatGPT ask for login via google. after connecting, website ask for mobile number verification but entering any number website giving error. he tried 4-5 times but same issue and not resolved yet. I signed up months ago and it's working fine for me.

LoyceV-When the bots become better, we will need to change our behavior. And the main part of this change in behavior is learning what reversible computation is.

It will not take long time and every one will accept bot as a information provider. Many Youtubers and free lancers are using Chatgpt and without knowledge of specific skills they are making quality videos, article, programming code and solving other problem. This is just start like BTC was in 2009. In five year i believe we will change our behavior.

[jvanname]

Gladitorcomeback-If people need GPT to produce text, then maybe they should not be communicating at all. That is absolutely horrible. And I bet that this applies to about 90% of all communication in certain fields. After all, there are many people talking about AI these days, but hardly any of them know about reversible computing because they are horrible.

[Marykeller]

What I like about AI ChatGPT is the way it breaks down questions into simple and understanding terms that a dullard would easily comprehend.

My findings about ChatGPT are that it gives different answers to questions being asked. The answer he gives to you, won't be the same answer it gives to another person when asked. There are different piles of answers to a particular question.

I tried to ask the same questions you asked AI ChatGPT, it wasn't written the same way as yours but if you read through, you will find out that the explanations are the same.

[cygan]

Demis Hassabis, the ceo of DeepMind (a subsidiary of google) says that their new ki Gemini will be more powerful than OpenAI's ChatGPT.
as the wired.com article also states, Gemini will be based on the technology and techniques of AlphaGo, which was developed by DeepMind in 2016.

Google DeepMind’s CEO Says Its Next Algorithm Will Eclipse ChatGPT
Demis Hassabis says the company is working on a system called Gemini that will tap techniques that helped AlphaGo defeat a Go champion in 2016.

https://www.wired.com/story/google-deepmind-demis-hassabis-chatgpt/