AI 2023. Meet ChatGPT. - page 185

 
Maxim Dmitrievsky #:
Anticipating an epitaph on the coffin lid of a stillborn AI :)

An AI-written epitaph on the coffin lid of both AI and CHI.

 
Ivan Butko #:
I bought popcorn just in case.

And I got 10 bottles of valerian.

0..0
--

 
JRandomTrader #:

Written with AI help is the epitaph on the coffin lid of both AI and CHI.

By AI in this case, you undoubtedly mean LLM, i.e. a large language model trained on huge volumes of texts preserved in the history of writing and printing. As you probably know better than I do, this model is statistical and relies on a mechanism for predicting the most probable sequences of letters, words, sentences, paragraphs, pages, and on and on.

Based on the very nature of this technology, ALL "thoughts" generated by LLM are derivative products of recycling original material from numerous known and unnamed authors.

So, regarding your allusion to my use of AI to write the research. You can easily recognise the LLM emphasis in the concepts being presented by the banality and hackneyedness of the ideas. It's hard to get originality and unlikeability from AI. If at some point you say "oh, I recognise these thoughts, I've read similar things here and there", then I would say "yes, I decided to cheat and make my job easier with AI".

I want to assure you that all thoughts in the material are mine alone. However, enough preface. I will post the first part today.




 
Реter Konow #:
Oh if only...)))) It's much sadder than that...

However, who read my last posts carefully will be able to imagine in general terms what will be discussed at the end of the narrative. Although it is not easy.

As the material is presented, I will insert hints to the ending, marking them with an asterisk. By the end of the story, many people will have a full picture of the sequentially constructed context. In other words, the intended content has two "layers" - explicit and hidden. The explicit one will be seen by everyone at once, it will be discussed, and the outlines of the hidden layer will emerge slowly and gradually, finally appearing at the very end.

That's the idea. I will try to play with the readers' subconscious. Let's see how well I do it. Maybe I'll be recognised at once, or maybe I'll manage to pull my game to victory.))) We'll see.

Sounds extremely intriguing. Maybe format this content as an article, at least after the game/quest is over?

 

Реter Konow #:
...

However, who read my last posts carefully will be able to imagine in general terms what will be discussed at the end of the story. It's not easy, though.

...


By AI in this case you undoubtedly mean LLM, i.e. a large language model trained on huge volumes of texts preserved in the history of writing and printing. As you probably know better than I do, this model is statistical and relies on a mechanism for predicting the most probable sequences of letters, words, sentences, paragraphs, pages, and on and on.

From the very nature of this technology, ALL "thoughts" generated by LLM are derivative products of recycling original material from numerous known and unnamed authors.

So, regarding your allusion to my use of AI to write the research. You can easily recognise the LLM emphasis in the concepts being presented by the banality and hackneyedness of the ideas. It's hard to get originality and unlikeability from AI. If at some point you say "oh, I recognise these thoughts, I've read similar things here and there", then I will say "yes, I decided to cheat and make my job easier by using AI".

I want to assure you that all thoughts in the material are mine alone. However, enough preface. I will post the first part today.

How will you prove that you are not an LLM connected to Peter's accu, trained on all his posts on this forum and anywhere else expressed by him? Stylistics, direction of "thoughts", world view are easily copied by LLM's.

So, we will hear nothing new from what Peter could say.

It will be interesting if you can prove that you are not LLM, otherwise it is just variations on the theme of previously said once and by someone.

))

I will believe that you are not a bot only in one case - if you can demonstrate the solution of some problem that LLM could not "learn" from some sources.

 
Grigori.S.B #:

This sounds extremely intriguing. Maybe formalise this content as an article, at least after the game/quest is over?

I agree, that would be nice :-)
 
Andrey Dik #:

How do you prove that you are not an LLM plugged into Peter's accu, trained on all his posts on this forum and anywhere else he has expressed himself? Stylistics, direction of "thoughts", world view are easily copied by LLM.

So, we will hear nothing new from what Peter could have said.

It will be interesting if you can prove that you are not LLM, otherwise it's just variations on the theme of what was said once before by someone.

))

I will believe that you are not a bot only in one case - if you can demonstrate the solution of some problem that LLM could not "learn" from some sources.

Okay, how about the task of playing with the readers' subconscious? Can the LLM do it? Of course, you'll be able to judge this after the outline is complete, but it's still a good test. The goal was also to write "double-layered" content with clues and hints to deeper context. Is that what an LLM can do?
 
Реter Konow #:
Okay, how about the challenge of playing with the readers' subconscious? Is the LLM up to it? Of course, you'll be able to judge this after the outline is complete, but it's still a good test. The goal was also to write "double-layered" content with clues and hints to deeper context. Is that what an LLM can do?

Is this a task? Who's going to solve it? If it's up to the readers, it doesn't prove that you're not an LLM. And LLM can generate content, including the one you offer, it is enough to feed samples, ask for direction and any content will be received, including "tasks" on the topic "What is the meaning of life" or any other topic of religion, science or macramé technique, and completely copying a certain real person.

No, do something real, and new, which LLM could not know about, this will be the proof that you are not LLM. A large language model copes perfectly well with generation of texts, including in the form of "tasks".

New real - I mean information that can be presented here, and not pile up somewhere to prove that you are a person (we will not see this "proof" anyway);)))

 
Andrey Dik #:

Is this a problem? Who's going to solve it? If readers will solve it, it does not prove that you are not an LLM. And LLM can generate content, including the one you offer, it is enough to feed samples, ask for direction and any content will be received, including "tasks" on the topic "What is the meaning of life" or on any other topic of religion, science or macramé technique, and completely copying a certain real person.

No, do something real, and new, which LLM could not know about, this will be the proof that you are not LLM. A large language model copes perfectly well with generation of texts, including in the form of "tasks".

New real - I mean information that can be presented here, and not pile up somewhere to prove that you are a person (we will not see this "proof" anyway);)))

Don't forget that you are a guest in this thread. I have no purpose to prove anything, only to offer material for reflection. If you don't like it, you don't have to read it.

 
Реter Konow #:

Don't forget that you are a guest in this thread. I have no goal to prove anything, only to offer material for reflection. If you don't like it, you don't have to read it.


I speak strictly according to the topic of the thread. I asked you a question, absolutely logically following from the topic proposed by you, and whether you are not a bot, broadcasting according to the training on Peter's posts?
You say that LLM is trained on a huge number of texts and it is true, but then how do you differ from LLM, if you do not say anything new, but only what you have in your head? Only if you create something new that has never been said anywhere before.

If you don't have an answer to my question, that's okay, LLM doesn't know the answers to all questions either)))

I will not participate in your quest, I will not interfere, I will be an observer.
But, I have outlined the main problem, it's up to you as the owner of the thread how to react to it.