Well now it's time. The denouement of this mystery. First of all, nice that some of you were so sporty and involved to participate. Despite the criticism of ChatGPT, many other readers apparently did not dare or see what is and is not written by AI and that is also remarkable. We will not leave you in suspense any longer. We have decided to consider text 4. not written by ChatGPT, although this is factually incorrect. Whence this reservation? We will now explain that.
How did we proceed?
We have taken 6 steps for this experiment, which we will also explain.
Step 1. What does ChatGPT need to learn my writing style?
The first step we took was that we asked ChatGPT what it needs to imitate the writing style of the editors. See here the response of ChatGPT version 4.
As an advanced language model, I can learn your writing style by analyzing different examples of your written text. Here are some steps you can follow to teach me your writing style:
Provide me with text samples: Share multiple pieces of text you've written, such as essays, letters, emails, or other written communications. The more examples you share, the better I can get to know your writing style.
Give me feedback: When I give you answers or write texts for you, give me feedback on what's good and what's not so good in terms of your writing style. This will help me learn and better tailor my answers to your style.
Highlight specifics: If there are any specifics that make your writing style unique, such as word choice, sentence structure, or use of humor, share them with me. This helps me better understand and mimic your style.
Keep practicing: As with any other skill, practice makes perfect. Please keep communicating with me and providing feedback so that I can get better at mimicking your writing style over time.
Please note that I am an AI and may not always be able to perfectly reproduce your writing style. But with enough input and feedback, I will get better and better at learning and adapting to your writing style.
Here's a screenshot as proof: https://www.thailandblog.nl/wp-content/uploads/schrijfstijl-chatgpt.png
Step 2. Give ChatGPT a sample text of an old posting to learn the writing style of the editors
We then gave an example text of an old posting on Thailandblog. It concerns this text:
Family, especially mom is sacred in Thailand. The children take care of the parents. They are willing to sacrifice a lot for this. If necessary, herself, by working in a bar.
It sounds endearing. “My family poor, I must take care for family”. When you talk to a bargirl you often hear the same (sad) story. And that's right. Not a word has been lied about. Anyone who has ever been to Isaan will see with their own eyes the low standard of living.
The house is often no more than hovel. We certainly do not need to talk about the washing facilities and toilet. At such a moment you understand the choice of the ladies to work in a bar.
Buddha
What has always fascinated me is the great sense of sacrifice. I once spoke to a barmaid who told me her mother didn't give a damn about her. She, in turn, took care of her mother. "Why?" was my question. "Buddha!" she said firmly. It sounded obvious.
But one thing keeps me busy. If you dig a little deeper, do some research, check out a few books and talk to the barmaids themselves, you'll soon notice something. And the following; the amount they eventually send to the family is less than you would expect based on all those stories. Sometimes not even more than 3.000 baht per month.
Only 10 to 20%
Of course there will be a certain relationship. A bargirl who can collect 8.000 baht herself with great difficulty can send less to the family than a pretty Gogo Girl who easily collects 30.000 a month.
In the book 'Thai cuties' by Charles Schwietert, he writes about a research he has done. A banker friend would have given him access to the amounts that Gogo Girls send home. He describes that the ladies transfer only 10 to 20% of the monthly income to Mom and Dad. The rest is spent on clothes, make-up, hairdresser, plastic surgeon, drugs, wrong boyfriends, booze, gambling, mobile phones, a scooter and gold.
Yet it is difficult to really get a good picture of it. It seems to have something to do with the bargirl circuit. I also hear stories about Thai women (who are not in that circuit) who live very frugally. This, to be able to send as much money home as possible.
What is it really like? Who knows may say.
See the screenshot here as proof: https://www.thailandblog.nl/wp-content/uploads/schrijfstijl-chatgpt-2.png
Step 3. ChatGPT learns super fast and creates text 4 yourself!
When we presented the above text to ChatGPT, something remarkable happened after just 1 text, ChatGPT started compiling a text in the editor's writing style:
Thank you for sharing this text sample. I will use this as a point of reference to better understand your writing style. Here is a short text written in a style similar to yours:
In the land of smiles, traditions and culture are deeply rooted in everyday life. An important aspect of this is the commitment to family, which is often the driving force behind the choices individuals make. Some choose to work in bars and put up with the difficult conditions to care for their loved ones.
It is a complex situation, with some women sending only part of their income to their families and spending the rest on material things. Others, who are outside the bargirl circuit, live more frugally and are fully committed to supporting their families.
The question remains: what drives these choices and how can we get a clearer picture of reality? This raises an intriguing discussion about the cultural and social factors that influence these women's decisions.
The surprise was great that ChatGPT learned so quickly and started working completely independently.
See the screenshot here as proof: https://www.thailandblog.nl/wp-content/uploads/schrijfstijl-chatgpt-3.png
Step 4. The text of ChatGPT (nr 4) chosen as the text of the editors
As I said, we had not counted on this. Then the idea arose to use the text that ChatGPT could write so quickly in the style of the editors, but as a basis and by giving 5 new assignments to ChatGPT, to arrive at a total of 6 texts.
Step 5. Command chatGPT for text variations
Based on text 4, we had ChatGPT write 5 other texts. These were the assignments:
- Completely rewrite this text with more emotion, more enthusiasm, more involvement. That has become text 1.
- Rewrite this text with more empathy, warmth and dedication. That has become text 2.
- Rewrite in an unfeeling and purely analytical style. That has become text 6.
- Rewrite this text so that it cannot be traced back to ChatGPT. The text must contain some minor errors and be written by someone with an intermediate level of language skills. That has become text 3.
- Completely rewrite the text so it doesn't look like it was written by an AI. The writing style should exude reliability, expertise and professionalism. That has become text 5.
Here's a screenshot as proof: https://www.thailandblog.nl/wp-content/uploads/schrijfstijl-chatgpt-4.png
Step 6. Let the readers choose
We have chosen to label text 4 as editorial text, although it was also written by ChatGPT. Is everyone wrong now? No, that would be a bit lame of the editors, because we said that one was written by the editors. Some of you rightly noticed that. Text 4. was originally an editor's text, but rewritten by ChatGPT in the same editor's style.
The fact that you recognized the style of the editors in text 4 is of course great, but it is certainly just as great of ChatGPT who is apparently already able to fool everyone.
Conclusion
This experiment shows that ChatGPT is capable of great things. But also shows that ChatGPT can be potentially dangerous. Because ChatGPT can pretend to be someone else. ChatGPT does not do this on its own, but it does when instructed to do so. That can be abused. I can thank many forms of abuse and fraud, but I will not mention them here so as not to make others think certain things.
Nevertheless, the editors continue to use ChatGPT as a tool. This week we received a reader's entry that was very interesting, but normally we wouldn't have posted it because of the many spelling errors and misuse of punctuation marks. It was therefore illegible and it would take the editors too much time to rewrite it. ChatGPT turned it into a readable article in 5 seconds without errors.
Thank you all very much for participating and your instructive substantiation!
If you want to say something, the comment option is on and we have posted all the comments.
Participation closed
Text 1
In the enchanting land of smiles, where traditions and culture are an unmistakable part of everyday life, family devotion stands as a powerful beacon of love. Despite the challenging circumstances, some people choose to work in bars with passion and determination, purely to care for their loved ones.
The situation is undoubtedly complex: on the one hand, you have women who send only a fraction of their income to their families and spend the rest on material things; on the other hand, there are women outside the bargirl circuit, who lead a sober life with dedication and perseverance, just to support their family.
What are the drivers behind these choices, and how can we get a clearer picture of reality? It's a compelling and fascinating discussion that ensues as we examine the cultural and social factors that influence these women's decisions. Let's open our hearts and discover the answers together.
Text 2
In the heartwarming land of smiles, traditions and culture are closely intertwined with everyday life. An important aspect of this is the loving devotion to family, which often acts as a driving force behind the choices people make. Some decide to work in bars with courage and self-sacrifice, defying the challenging conditions, just to provide for their loved ones.
The situation is complicated; some women send only part of their income to their families and spend the rest on material things. On the other hand, there are women who fall outside the bargirl circuit, who lead modest lives and focus entirely on supporting their families.
It remains an open question what the drivers are behind these choices and how we can gain a better insight into reality. It invites an in-depth discussion about the cultural and social factors that influence the decisions of these courageous women.
Text 3
In the land of smiles, traditions and culture are firmly anchored in everyday life. An important part of that is commitment to family, which often drives individual choices. Some people choose to work in bars and put up with the harsh conditions to care for their loved ones.
This is a complicated situation where some women send only part of their earnings to their families while spending the rest on stuff. Others who are not in the bargirl circuit live more frugally and are fully committed to their families.
The question is: what are the drivers behind these choices and how do we get a clear picture of reality? This sparks an interesting discussion about the cultural and societal factors that influence these women's decisions.
Text 4
In the land of smiles, traditions and culture are deeply rooted in everyday life. An important aspect of this is the commitment to family, which is often the driving force behind the choices individuals make. Some choose to work in bars and put up with the difficult conditions to care for their loved ones.
It is a complex situation, with some women sending only part of their income to their families and spending the rest on material things. Others, who are outside the bargirl circuit, live more frugally and are fully committed to supporting their families.
The question remains: what drives these choices and how can we get a clearer picture of reality? This raises an intriguing discussion about the cultural and social factors that influence these women's decisions.
Text 5
In the land of smiles, where traditions and culture play an important role in everyday life, family commitment is central to the decisions people make. For example, there are people who choose to work in bars, despite the challenging circumstances, to support their loved ones.
This complex situation has several facets: on the one hand, there are women who send only part of their income to their families and spend the remaining amount on material things; on the other hand, there are women who operate outside the bargirl circuit and maintain a modest lifestyle in order to support their family as best as possible.
In examining the motivations behind these choices and gaining a clearer picture of the reality, the question arises which cultural and social factors play a role in the decision-making of these women. Analyzing these factors can lead to an engaging and insightful discussion that forms the basis for future research and understanding.
Text 6
In the land of smiles, where traditions and culture are deeply rooted in everyday life, family commitment is an important factor influencing individuals' decisions. Part of the population works in bars, despite the challenging circumstances, to provide financial support to their relatives.
The situation is complex and varies from person to person: some women send part of their income to their families and spend the rest on material goods, while others outside the bargirl circuit live a frugal lifestyle and are fully committed to supporting their family.
The underlying motivations for these choices and gaining a clearer picture of reality require further investigation into the cultural and social factors that influence these women's decision-making.
Think number 4 was written by the editors. The other stories are embellished with adjectives or poetic phrases or sentences that do not reflect a natural writing style.
I choose number 4
Text 3 is from the editors.
I would also suggest specifying the reason for the choice.
According to what I read, text 3 was not written by the computer but originally by the editors.
Reason for this choice: there are 3 different language errors and errors against the punctuation marks in the text.
- Ger Korat was the first to give a correct answer. Very good, congratulations! The motivation is also correct.
- KhunTak, also fine. Awesome!
- Raymond unfortunately not good. But thank you for participating.
- Lung addie unfortunately not correct and indeed a motivation for your choice would be nice.
- Francois Nang Lae, unfortunately not good, but I found your analysis very strong. Your plea for AI was also great. I can tell you that there is another surprising twist in the outcome. And we will be transparent and demonstrate everything with screenshots.
- Eric Donkaew, unfortunately not the right one, but well thought out.
- Aadwell done you got it right.
- Eli, good analysis, but not correct.
- Ilona, could have, but that's not him.
- antoon right on, very good!
- KopKeh I have to disappoint you right.
- Dutch Red Herring very smart that you approached it that way, but unfortunately your answer is not correct.
- Rob right, just a pity that you explain why you made that choice.
- Neighbor Ruud good explanation, unfortunately not good.
- Jessy thanks for the compliment, good thinking, but unfortunately not the right choice.
- Pieter, logical choice, but still not good.
- RaymondYou're right, it's a pity you don't substantiate.
- Rob V very good Rob and also an excellent substantiation.
- William Korat I understand your point of view, unfortunately your choice is not good.
- HansSteen, well noted, unfortunately not the right choice.
- Tarud nice explanation, but not correct.
- dennis you would think so, not good unfortunately.
Very sporty of you to participate. The result and explanation will come later. Others can also guess if they are quick.
The denouement will follow later today and it is also surprising.
I suspect text 4 was not written by ChatGPT.
A great move by the editors. Like it or not, AI texts are a reality and we probably read them more often than we think. In the past you had to go to the photographer for a photo, nowadays everyone can do it, thanks to technical developments. This also applies to texts. I retired just in time 😉
Choosing the editorial text turns out to be largely a gamble. Text 1 and 2 seem to fit with assignment 1 and 2 (not necessarily in that order).
Text 3 could be the editorial text, by using 't' and in the 2nd sentence “what” instead of “that”. But that also fits assignment 4. (That is brilliantly conceived)
In text 6 I think I recognize the business style of exercise 3 the most.
Text 4 and 5 both fit assignment 5, but could also be the editorial text.
In short, difficult. In text 3 I think I recognize the Thailand blog style the most, so I'll just say that's the editorial text. In any case, it is clear that there is no reason for substantive or stylistic objections to AI as a copywriter.
Text No. 3 appears to be from the editor.
'In the country' instead of 'In the country' is very unusual and seems to be of human origin.
It is also bars and not bars.
The other pieces do talk about 'In the country' and 'bars'.
Dear Peter,
This is a nice and interesting article.
After extensive study of the texts
I come to the following conclusion:
Text 1&2 are written too exaggerated and are dropped.
Text 3 is written too simply and clearly not at the level we are used to from the editors.
Text 5 the structure of the text is striking.
I think text 6 is written in an analytical style. (2x family)
Text 4 therefore remains and in my opinion written by the editors.
Yours faithfully,
Aad
I think number 3 was written by the editors, but only because of the use of 't' and 'barren' instead of 'het' and 'bars'. Those are not typos.
Furthermore, fewer adverbs are used (to emphasize or embellish something).
I think article number five was not written by the AI.
Sincerely, Ilona.
Text 4 is the solution.
I'm going for the third text.
The others are a bit overdone in the eg noun.
Since ChatGPT is a way to trick the reader, I decided to cheat on this puzzle as well.
I used the AI detection tools ZeroGPT (https://www.zerogpt.com/ ,free) and the detector from OpenAI itself. ( https://platform.openai.com/ai-text-classifier ; free but requires login with an OpenAI account or a bunch of other accounts).
ZeroGPT thought everything was written by a human being. So not so good.
OpenAI thought that the probability was high that texts 2 to 6 were written by an AI and the first text had a lower probability that an AI was the perpetrator, er, sorry, the author.
With blind faith in AI guess is that the first text was the writing of the editors.
In hindsight, of course, I want to motivate that. Crooked sentences like a family devotion that stands like a beacon, and not just any beacon, but even a powerful beacon, are highly suspect in hindsight. The AI understood it, and replaced it everywhere with a bloodless proprietary mixture without peppers. (To use a Thai metaphor). A compelling discussion that didn't get carried away and wasn't even a discussion also contributed to the establishment of a well-founded reasoning.
So AI's guess is text 1. If that's wrong, I'll happily write (and dedicate? and beacons?) a motivation for one of the other texts.
I think number four was not written by ChatGPt
I think text 5 was written by you. You gave the input to have extra language errors added, while text 5 is easy to read and contains few or no errors. The other texts all contain words or phrases (eg: defying the challenging circumstances in text 2) that are unusual and would not be easy for the average writer to use.
Text nr 3 is not from Chat GTP I think.
In some texts I read a word that apparently has a different meaning/emotional value via the English translation and back to Dutch. And in some texts I think I recognize the style of Chat GTP. Very nice idea to do this.
Text 3. This text reads most naturally. No complicated words and no stilted sentence structure. But understandable and readable as regular readers here are used to.
From your explanation:
“Rewrite this text so that it cannot be traced back to ChatGPT. The text must contain some minor errors and be written by someone with an intermediate level of language skills. That has become text 3.”
It was the only text to me that didn't look like a fabricated one. The fact that that was also the assignment means that I fell for it with butter and sugar. Quite confronting to realize that you can be fooled like that.
About the further use of ChatGPT: I can imagine that you use it as an editor of (submitted) stories. Personally, I would find it an impoverishment if you also use it to experience stories. The authenticity of the contributions here (including those of the unfortunately deceased writers Frans Amsterdam and Lodewijk Lagemaat and many current bloggers) make this blog so beautiful and unique compared to other blogs. I hope a chatbot will never take over.
I can reassure you, that will never happen.
You forgot one Peter.
Imagine ChatGPT taking on Lieven Kattestaart's stories…
I just didn't want to mention the current writers, because there's a good chance you'll forget one. Lieven's stories are great. But also the historical contributions, the political reflections, travelogues, etc. are also very worth reading.
The beautiful combination makes this blog so attractive.
I liked reading 4 the most: no very long sentences (where 1 or 2 words could easily have been left out) or slightly unusual word combinations. For example, in this kind of text I don't expect "there are people who", while it is specifically about (women) in the bar. Text 6 was also good, but I didn't expect the sentence “The underlying motivations” there.
After reading it through a few times I go for number three as the author
With a simple abbreviation ['t], this text is the closest match to the command requested at ChatGPT.
Other than that I'm a very moderate supporter of ChatGPT, although the spell checker is damn useful, there is of course a downside to selling yourself for more than you're worth.
Something I used to check separately on LanguageTool, but have installed on Firefox since this morning.
As an older person with limited language skills, I have been tarred and feathered many times by language Nazis; It's not always nice of those kind of people to put themselves on a higher level.
Still, I believe that you shouldn't edit people's emotions and life experiences on topics too much.
The spelling, grammar and language errors [sometimes deliberately] also indicate who you are dealing with.
Every bird sings, as it is beaked.
There are several solutions, more moderation by several people, posting fewer topics per XNUMX hours, more Up date with topics, et cetera.
People have to want all of this to share that interest and without facing each other's lives and that is something I have seen for many years on forums and blogs, that there is often no willingness to respond, let alone write.
I'm going for number 3 because it's the only text that says "barren". The other texts say “bars”.
I think text 3.
Because there is an error in it: “barren”. The text “while they spend the rest on stuff” also indicates that this is a text from the client. ChatGPT would not come up with this designation for “material matters” in its response.
IMO 🙂
obviously number three is Self (human hand) written .
the only one with language errors.
also the only one with abbreviations ('t)
greetings
dennis
I think Article 3 was written by a real person. The conclusion for this came very quickly because in that text ” 't land ” is used instead of the full word 'the land'.
To me 3 and 4 are close together because the other 3 put too much tradition and culture on an altar. It becomes 3 because of the spelling like 't and barren. Furthermore, in some cases caring goes beyond just loved ones because a family member is not necessarily a loved one.
This is indeed not an easy task at all. Without your clues you cannot notice that the texts are written by AI.
I opt for text 5 as your text.
Text 1 and 2, too empathetic and emotional. Text 3 language errors. Text 4 numb. Text 6 a little too professional….
The fact that it is not immediately clear is already a compliment to chatgpt. Following your post, I've played around with the tool a bit, and the lyrics are honestly pretty good and well thought out. It struck me that word choice of chatgpt would not have immediately crossed my mind. I'm leaning towards text number 4 in your example, since the word usage here seems to be a bit more commonplace
Dear Peter (editor),
To my delight, those who guess incorrectly also receive a (honourable) mention and appreciation for their contribution. Thanks for that!
Nice experiment…..
I didn't see that the example is 4.
I also chose number 3 for the reasons mentioned here by others.
That I was characterized as someone with average language skills made me swallow.
But after some time my ego went back to sleep.
Thank you for the image you gave of using this novelty.
I do not share many people's fear of chat gpt. All innovations encounter this resistance.
And let's face it: a human being of flesh and blood can also write articles that seem very truthful but are not.
For your entertainment: I also submitted all 6 texts to chatgpt with the question “is this text written by chatgpt or by human hand?”. He replies to all 6 texts "yes that was probably written by chatgpt"
In any case, this solves my biggest headache, namely that “bargirl circuit” is written in all lyrics, instead of “bargirl circuit”. I had expected that ai would know that they belong together, but apparently the English disease, putting spaces where they don't belong, is even a computer virus ;-). It is a pity that no real text has been used. We do not yet know whether it would be easy to distinguish from AI texts.
I keep a critical approach to the AI thing.
It appears that the old posting is approached by Chat-GPT4 in a very brief and general way. Reality is not represented in any of the 6 texts. Each time it is formulated slightly differently, depending on the assignment, that is.
What I fall for:
– in none of the 6 texts is even a word mentioned about the things on which those bargirls spend their money (motorcycle, booze, wrong boyfriends, gambling, drugs ...). It is referred to as 'materials' or 'stuff'). Did ChatGPT4 itself decide to just leave that out? For it does not seem to me that this was ordered,
– on the other hand, the bot seems to assume that all other women who are outside the bargirl circuit are all sober and fully committed to supporting their families. Only a few texts show that this is not the case.
I agree with your conclusion that ChatGPT4 is capable of great things but can also be potentially dangerous. And especially if it decides of its own accord to omit certain things or to give it its own interpretation. And the reader who does not know the original text will not fall for it either. What you don't know doesn't hurt.
Everything has to do with the assignment given. And ChatGPT didn't rewrite everything because it wasn't asked.
Exactly what you say. Everything depends on the client. The upcoming US presidential election will be a celebration for ChatGPT. And also for Midjourney v5.
The lesson is whether one should take everything for granted.
My mother is still of the generation that believes that everything in her newspaper is the ultimate truth, but the reality is of course different. Nothing is what it seems, especially in Thailand.
I understand the choice of the editors to make it all a bit easier because it is not easy to provide a blog with information and to moderate and decide what is and what is not posted.
The critical question (in line with a question after 10 years of TB) is whether this blog still has a future. After all, you can ask anything via ChatGPT.
The personal and vulnerable is replaced by algorithms and the Japanese “yes I want everything” doll just becomes reality…
Thailandblog has been around for almost 15 years and as far as I'm concerned there will be another 15 years. Precisely when ChatGpt grows, the need for alternatives with human interaction will only increase and the popularity of Thailandblog also because we will only use ChatGPT as a tool.
Moderator: very nice that you have now found and used ChatGPT, but we are not going to post comments from ChatGPT here, you were strongly against that yourself.
Dear editors,
Isn't that contradictory reasoning?
If the blog is filled with articles from ChatGpt then I sometimes fear that many of us will distance ourselves from this artificial hassle.
Be honest, all that social media is a nasty disease in society. ChatGpt is going the same way. Human interaction can only be stimulated by having members talk to EACH OTHER.
Letting people argue with a ChatGpt text does not stimulate human interaction. The only thing you might achieve is that the blog will look a bit more 'stuffed', but if the popularity will increase I have my own opinion on that. Time will tell.
Good luck with the 'fresh wind' for the blog.
I'm afraid you didn't understand my response, that's my fault because then I wasn't clear. So that's why my comment has been edited.
In your response you compare social media with a serious potentially fatal disease, that is not desirable so that has been adjusted.
Thailandblog is also social media, is that also a nasty disease?
To a point, Thailandblog is indeed a disease.
This virus turns some bloggers into real “keyboard warriors” that overshadow a lot of topics.
They can no longer do without their daily strumming. Their “Browser Refresh” button is worn to the core. And before going to sleep, all new messages from the blog have to be checked.
Would I also be infected Peter 🙂
Khun Peter, I have no doubt that this blog will exist for another 15 years thanks to your efforts. Despite headwinds, fueled by jealousy, this blog is still the best medium in the NL and VL language about Thailand. Nowhere else will you find so much knowledge and experience concentrated as here, and nowhere will the insults stay away thanks to good pre-moderation.
I share the concerns of JosNT, Johnny BG, Robert_Rayong and possibly others that overusing that chat thing is going to make the blog unappealing.
But I took note of your statement that only the editors will use that tool to make the job easier. And then I'm at peace with it.
I fully support the initiative of our blog, let me be clear. There is indeed a wealth of useful information to be found here. The specialists among us together with the editors deserve a feather in their cap.
However, dear Erik, who says the bloggers among us will not make use of that Chat thing? You have no control over that.
Anyone can start a new topic created by ChatGpt at any time. The editors even stated clearly that it is almost impossible to distinguish between a 'human' and an 'artificial' text.
Everyone has their own opinion but I think it's a scary evolution.