English is not my primary languge. I am generally proficient but not quite on native speaker level yet. I'm pretty sure there are a lot of people in similar position as mine. AI is getting better and better, and is becoming increasingly common. I think that it would be nice to decide more preciecly what is and what isn't allowed. IMO it would be best to mark reviews with tags like:
AI revised
AI translated
auto-translated
Or at least put in FAQs more precise guidelines and add a link to FAQs in Rewiew rules. I personaly use AI on daily basis to redact pretty much every comment I am puting into internet and i see great value in that but i also wouldn't be suprise if someone would prefer raw, unfiltred rewievs.
-----
General conclusion after ~20 replies:
It is better to fully write review in Your native language and use dedicated tool for translation (DeepL) than partialy/fully write review and use LLM (ChatGPT) to translate.
With LLM it might get flagged as ai-written and it less accurately reflects initial vision.
English is not my primary languge. I am generally proficient but not quite on native speaker level yet. I'm pretty sure there are a lot of people in similar position as mine. AI is getting better and better, and is becoming increasingly common. I think that it would be nice to decide more preciecly what is and what isn't allowed. IMO it would be best to mark reviews with tags like:
AI redacted
AI translated
auto-translated
Or at least put in FAQs more precise guidelines and add a link to FAQs in Rewiew rules. I personaly use AI on daily basis to redact pretty much every comment I am puting into internet and i see great value in that but i also wouldn't be suprise if someone would prefer raw, unfiltred rewievs.
What they don't want is AI written reviews. Using AI to translate your own words from your primary language into English is fine. Using an app to help you edit your review is fine.
They don't want people - of any language - to simply tell an AI to write a review about a game that they think has good art, bad writing, etc...
If English is not your primary language, you are welcome to use an app to translate your review that you wrote in your primary language. In my opinion, it would be best for people to use such an app when they believe aren't proficient enough in English to write clearly. You can still learn writing English better while using the app.
What they don't want is AI written reviews. Using AI to translate your own words from your primary language into English is fine. Using an app to help you edit your review is fine.
They don't want people - of any language - to simply tell an AI to write a review about a game that they think has good art, bad writing, etc...
If English is not your primary language, you are welcome to use an app to translate your review that you wrote in your primary language. In my opinion, it would be best for people to use such an app when they believe aren't proficient enough in English to write clearly. You can still learn writing English better while using the app.
That is exactly what i expected, but IMO The literal meaning of the words differs from the expected approach. I would say that when i am using AI to redact a review it no longer is in my own words. Thanks for the answer.
That is exactly what i expected, but IMO The literal meaning of the words differs from the expected approach. I would say that when i am using AI to redact a review it no longer is in my own words. Thanks for the answer.
I have had pretty consistent results with deepl.com --- I've used it to take paragraphs from one language, translated them one-after-another into 4-5 different languages, and then translate it back to the first language with near exactness. Even when not exact, the meaning of the paragraphs did not change.
To be honest, this is a risky situation. While it is not forbidden to use AI to translate "your own words" into English, ChatGPT or other translation programmes often make texts sound robotic. Even though the words belong to you, after AI translation, it can seem as if the text was written by AI. I have no idea how to prevent this. (This text is currently being translated into English using DeepL)
Ai writen means asking ChatGTP to "Write me a 4 of 5 star review about an adult VN. The review needs to be between 200 and 300 characters and contain praises like "good written story" and "nice visuals" als well as how nice the big boobs of the female characters are."
And yes, the funny part about this is, that this prompt already has 231 characters, way above the required minimum.
That is exactly what i expected, but IMO The literal meaning of the words differs from the expected approach. I would say that when i am using AI to redact a review it no longer is in my own words. Thanks for the answer.
Kinda depents on the translator you use. I would suggest DeepL (like many here), there you can choose every word one by one by clicking on it and match what you want to tell or fit to your own writing style. I does work very well.
To be honest, this is a risky situation. While it is not forbidden to use AI to translate "your own words" into English, ChatGPT or other translation programmes often make texts sound robotic. Even though the words belong to you, after AI translation, it can seem as if the text was written by AI. I have no idea how to prevent this. (This text is currently being translated into English using DeepL)
As a person whose primary language is English and teaches composition and creative writing at the post-secondary level - that paragraph does not appear written by AI. I know you mention using ChatGPT for translation, but here you used DeepL --- I would be interested in checking whether ChatGPT translations trigger a warning.
Using a few apps to check:
It's a small writing sample, so it isn't necessarily extremely accurate (more text would help accuracy) but the detectors appear to agree with my initial reaction.
Ai writen means asking ChatGTP to "Write me a 4 of 5 star review about an adult VN. The review needs to be between 200 and 300 characters and contain praises like "good written story" and "nice visuals" als well as how nice the big boobs of the female characters are."
And yes, the funny part about this is, that this prompt already has 231 characters, way above the required minimum.
Kinda depents on the translator you use. I would suggest DeepL (like many here), there you can choose every word one by one by clicking on it and match what you want to tell or fit to your own writing style. I does work very well.
DeepL is great for translation, but I'd rather use Gemini and ascertain that it didn't mess anything up. It gives better results for someone on a ~C1 English level. But if someone is weaker, then DeepL is probably a better choice.
As an English instructor at the post-secondary level - I disagree. But you do you. Already, I believe I have pointed to one error with Gemini's translation.
As a person whose primary language is English and teaches composition and creative writing at the post-secondary level - that paragraph does not appear written by AI. I know you mention using ChatGPT for translation, but here you used DeepL --- I would be interested in checking whether ChatGPT translations trigger a warning.
It's a small writing sample, so it isn't necessarily extremely accurate (more text would help accuracy) but the detectors appear to agree with my initial reaction.
I'm glad to hear that. I haven't written any game reviews yet, but I don't want to have any problems with that. I mostly use ChatGPT for longer texts. For quick messages like this, I prefer Deepl. (Ctrl + C and text translated)
DeepL is great for translation, but I'd rather use Gemini and ascertain that it didn't mess anything up. It gives better results for someone on a ~C1 English level. But if someone is weaker, then DeepL is probably a better choice.
To follow up - I asked Gemini: "For translating personal written works into English, between Gemini AI and DeepL, which application is considered better? Create a table comparing strengths, weaknesses, perceptions, critical reviews, and the like."
I'm glad to hear that. I haven't written any game reviews yet, but I don't want to have any problems with that. I mostly use ChatGPT for longer texts. For quick messages like this, I prefer Deepl. (Ctrl + C and text translated)
Since I was already doing one...
To follow up - I asked ChatGPT: "For translating personal written works into English, between ChatGPT and DeepL, which application is considered better? Create a table comparing strengths, weaknesses, perceptions, critical reviews, and the like."
yep revise is the word...
I don't think banning AI based on the scans is enforceable. It detected properly three paragraphs i added after revising with ai but it marked everything prior as likely AI generated despite me changing ~10% of text in those 13 sentences after Gemini revision.
DeepL is great for translation, but I'd rather use Gemini and ascertain that it didn't mess anything up. It gives better results for someone on a ~C1 English level. But if someone is weaker, then DeepL is probably a better choice.
I'd say it's the opposite. Letting an AI talk for you is way better for people with low language skills.
I am no native speaker either, but I am on a level of full english communication skill on a high level.
My only issue is, that most of the time I don't always find the best words and only the simpler ones come to mind.
And that's where this DeepL word-choosing is perfect. I can quickly construct what I want to say, but with better words.
While AIs like Gemini just writes their own thing with their own style.
I tested several LLMs in the past and none of them gave me satisfactory results, each of them sounded like it's written by a different person. It was simply not me anymore.
One time I reviewed something and later on another person reviewed it and it was just my own review fed into and hashed back out by ChatGPT.
I found it really bizarre. All my criticisms but half of it worded inaccurate to what actually happens in the VN, and all very GPT-y in style.
I don't think banning AI based on the scans is enforceable. It detected properly three paragraphs i added after revising with ai but it marked everything prior as likely AI generated despite me changing ~10% of text in those 13 sentences after Gemini revision.
I'd say it's the opposite. Letting an AI talk for you is way better for people with low language skills.
I am no native speaker either, but I am on a level of full english communication skill on a high level.
My only issue is, that most of the time I don't always find the best words and only the simpler ones come to mind.
And that's where this DeepL word-choosing is perfect. I can quickly construct what I want to say, but with better words.
While AIs like Gemini just writes their own thing with their own style.
...
...
I tested several LLMs in the past and none of them gave me satisfactory results, each of them sounded like it's written by a different person. It was simply not me anymore.
...
Fully agree with that i do not recomand using LLMs output as final text without validating it manually. And there always is something that seams a little bit robotic but it ain't that big of a deal in my opinion.
yep revise is the word... View attachment 5461546
I don't think banning AI based on the scans is enforceable. It detected properly three paragraphs i added after revising with ai but it marked everything prior as likely AI generated despite me changing ~10% of text in those 13 sentences after Gemini revision.
Do you know why it thought there was AI writing? I do - at least as an English instructor, I can tell you what is tripping the detector.
The composition:
The use of lists with colons. That's why. Humans, when we write, don't normally make lists like this (well, we do, when we are writing for academics, research papers, and the like) --- at least not in the English language. Even in reviews, it isn't necessarily that structured.
We write like this: with sentences, phrases, clauses, and the like.
Do we sometimes use bullets? Sure!
Do we sometimes use colons? I did.
Do we make lists? yes.
But these are the initial red flags for AI writing - headings, bullet points under those headings, short and precise reasoning. If I saw this from a student who turned it in as a writing assignment, I would be wary. Now, if I knew their writing style and it was similar to this - I wouldn't be as wary.
As for the content:
I am doubtful that you wrote this in your primary language and then had it translated to - "The narrative maintains a sensible level of suspense through well-placed mysteries" -- few people write this way, period. If that is how you are writing it in English (or close to it) or think that is how English is composed in today's world, I would suggest stepping back a bit and evaluating your learning.
"While characters face serious problems... simplified and superficial" feels like forced writing, not natural.
The entirety Illogical Plot Decisions point sounds forced, as well.
I would recommend, again, writing in your primary language and then using a better translation program, such as DeepL.
Fully agree with that i do not recomand using LLMs output as final text without validating it manually. And there always is something that seams a little bit robotic but it ain't that big of a deal in my opinion.
I also do not recommend without human eyes checking it over - however, I think you would best be served learning English from a better translation tool than attempting to write English and using an LLM to help you revise.
Well You raise a few good points depite beeing suprisingly inacurrate in Your diagnosis (edit: nevermind i just needed to reed a few more times to get Your point about the colons). I think I will switch into DeepL due to Your last argument about serving me better in learning english. I would be stiupid not to takean advice fromproffesional.
If You are curious i haven't used any tool in this thread i only looked up a few words.
ahaha - I would have fallen for the same trap, so now first thought "oh, maybe learned yet another way to detect German speakers", but as PStonix using Polish, that seems to be more of a multi-language different use of redact/or/ion (resp. Redakt/eur/ion).
Well You raise a few good points depite beeing suprisingly inacurrate in Your diagnosis (edit: nevermind i just needed to reed a few more times to get Your point about the colons). I think I will switch into DeepL due to Your last argument about serving me better in learning english. I would be stiupid not to takean advice fromproffesional.
If You are curious i haven't used any tool in this thread i only looked up a few words.
Wszystko, co jest zaznaczone na fioletowo, zostało przetłumaczone z języka angielskiego na polski przez DeepL.
Everything in Purple is DeepL translated from English to Polish.
Myślę, że dobrze sobie radzisz w nauce pisania po angielsku. Na podstawie porównania Gemini i DeepL uważam, że sztuczna inteligencja nie będzie tak pomocna w nauce i rozwijaniu umiejętności, jak w przypadku wyboru dedykowanego narzędzia tłumaczeniowego, takiego jak DeepL.
I think you are doing well in learning how to write using English. Based upon what I am seeing from the Gemini and DeepL comparison, the AI is not going to be as helpful with your learning and developing skills as it would be if you chose a dedicated translation tool such as DeepL.
Znalazłem przyczynę wykrywania AI:
I found what is causing the AI detection:
"chcę napisać review po angielsku po przez wymienienie silnych i słabych stron gry"
"I want to write a review in English by listing the strengths and weaknesses of the game:
"postaraj się przetłumaczyć moją wersję oraz jak najbardziej ją skrócić"
"try to translate my version and make it as short as possible"
Te dwa podpowiedzi nie nakazują sztucznej inteligencji jedynie przetłumaczyć Twoją pracę, ale nakazują jej stworzyć kompozycję w oparciu o pewne aspekty, które jej podałeś. To ona pisze Twoją recenzję. Nie piszesz jej Ty.
Those two prompts aren't telling the AI to just translate your work, you are telling the AI to generate the composition based upon some aspects you have given it to work from. It is writing your review. You aren't writing it.
Original
Gemini AI Generated
(edited by you)
English translation (DeepL)
jako silne element: grafika - modele i sceny engaiging - niewiadome w plocie trzymają w sensownym stopniu w napięciu, akcja nie rozwija się zbyt wolno i cały czas jest odpowiedni poziom napięcia
słabe strony:
płytkie postacie - postacie mają poważne problemy ale są one mocno zpłycone nielogiczność plotu - zachowania głównego bohatera są w dużej mierze nielogiczne a ponadto robi mnóstwo rzeczy nie tylko nie rozumiejąc dlaczego mu coś kazano ale równierz nie próbując dowiedzieć się dlaczego. Ma to coprawda sens w kontekście historii ale jest bardzo nieimersywne. interakcje między postaciami są bardzo ograniczone - główne NPC mają ze sąbą wzajemnie bardzo mało interakcji. Prawie zawsze mają sceny wyłącznie z głównym bohaterem system moralności z którego prawie nic nie wynika - plot robi z głównego bohatera bad boya i gra zamiast to wpełni (brakuje mi słowa po polsku embrace po angielsku) __ ocenia na bierząco moralność każdej z decyzji co prowadzi do kuriozum w pojedynczym przejściu możemy starać się iść grzeczną drogą po czym plot i tak robi z nas bad boya pętla rozgrywki jest nieprzyjemna - sprowadza się do ciągłego patrzenia w kalendarz jak i do mocno powtarzalnych aktywności
Graphics are worthgiving game a go but:
I recomand playing with some cheat mods, skipping activity dialogues (expect the ones You more visually pleasing) but everyting in reading story events.
Dr Sin really needs to team up with some decent writer and LewdLab might become one of the best team in industry, but right now is mid.
Strengths
Impresive graphics:models and sex scenes, really top-notch. Engaging Pacing & Plot: The narrative maintains a sensible level of suspense through well- placed mysteries. The action is neither too slow nor overwhelming, ensuring a continuous and appropriate level of tension.
Weaknesses
Shallow Characterization: While characters face serious problems, their development and depth often feel simplified and superficial. Illogical Plot Decisions: The protagonist's actions are frequently illogical and poorly
motivated. They perform numerous tasks without understanding or attempting to discover the why, which, despite having a rationale within the story's context, is highly
unimmersive. Limited Character Interaction: Key girls have very few mutual interactions; their scenes are almost exclusively with the main character. Flawed Morality System: The morality system feels largely consequential-less and
disconnected. The plot steers the protagonist towards a "bad boy" path, yet the game constantly judges every decision. This leads to the absurdity of trying to play a "good" route only for the main story to force protagonist to do "bad stuff". Unpleasant Gameplay Loop: The core gameplay is repetitive, revolving around constantly checking the in-game calendar and engaging in monotonous activities.
Graphics are worthgiving game a go but:
I recomand playing with some cheat mods, skipping activity dialogues (expect the ones You more visually pleasing) but everyting in reading story events.
Dr Sin really needs to team up with some decent writer and LewdLab might become one of the best team in industry, but right now is mid.
Strengths: Graphics - models and scenes Engaging - the unknowns in the plot keep you in suspense to a reasonable degree, the action does not develop too slowly, and there is an appropriate level of tension throughout
Weak points:
Shallow characters - the characters have serious problems, but they are very shallow illogical plot - the main character's behavior is largely illogical, and he does a lot of things not only without understanding why he was told to do them, but also without trying to find out why. This makes sense in the context of the story, but it is very non-immersive. Interactions between characters are very limited - the main NPCs have very little interaction with each other. They almost always have scenes exclusively with the main character. The morality system, which has almost no consequences - the plot makes the main character a bad boy and instead of fully embracing it, it evaluates the morality of each decision on an ongoing basis, which leads to a curious situation where in a single playthrough we can try to take the good path, but the plot will make us a bad boy anyway. the gameplay loop is unpleasant - it boils down to constantly looking at the calendar and highly repetitive activities
Graphics are worthgiving game a go but:
I recomand playing with some cheat mods, skipping activity dialogues (expect the ones You more visually pleasing) but everyting in reading story events.
Dr Sin really needs to team up with some decent writer and LewdLab might become one of the best team in industry, but right now is mid.
Oznaczyłem kolorami fragmenty napisane przez Ciebie (kolor zielony), wygenerowane przez Gemini (kolor żółty) oraz przetłumaczone przez DeepL (kolor niebieski). Fragmenty dodane/edytowane przez Ciebie do tekstu wygenerowanego przez Gemini pozostawiłem w kolorze białym i dodałem je do pozostałych dwóch.
Wszystko, co jest zaznaczone na żółto, to tekst wygenerowany przez sztuczną inteligencję, która ustaliła, co chcesz powiedzieć – a nie to, co faktycznie napisałeś w swoim tekście. To nie jest Twój tekst. To tekst sztucznej inteligencji. Słowa, które wybiera do generowania, oraz kolejność, w jakiej to robi, również budzą podejrzenia wykrywaczy sztucznej inteligencji.
Jak wspomniałem wcześniej, nie napisałeś „The narrative maintains a sensible level of suspense through well-placed mysteries” w języku polskim. Napisałeś „niewiadome w plocie trzymają w sensownym stopniu w napięciu”, co w języku angielskim oznacza „the unknowns in the plot keep you in suspense to a reasonable degree”. Chociaż wyrażają one podobne koncepcje, jedno z nich jest nie na miejscu i w ogóle nie pasuje do Twojego stylu pisania.
Wspomniałem również, że:
„While characters face serious problems... simplified and superficial” brzmi jak wymuszone sformułowanie, a nie naturalne.
Całość punktu „Nielogiczne decyzje fabularne” również brzmi wymuszona.
Przetłumaczona wersja „the characters have serious problems, but they are very shallow” brzmi bardziej naturalnie (choć brakuje jej jasności, ale tak samo było w oryginale).
Przetłumaczony punkt „Nielogiczna fabuła” również brzmi bardziej naturalnie.
Zabawne, że poprosiłeś Gemini o skrócenie tekstu, ale jest on najdłuższy z trzech (o 100 znaków dłuższy niż DeepL).
-----
Zalecałbym napisanie recenzji samodzielnie – w pełni i z zachowaniem poprawnej gramatyki, struktury i kompozycji języka polskiego – a następnie skorzystanie z aplikacji tłumaczeniowej, aby przetłumaczyć ją na język angielski. W ten sposób będzie to Twój tekst, Twoja recenzja i nie zostanie ona oznaczona przez żadne narzędzie wykrywające sztuczną inteligencję, jeśli takie zostanie użyte do sprawdzenia recenzji. Ponadto wersja tłumacza pomoże Ci lepiej wzmocnić swoje umiejętności pisania w języku angielskim.
I color-coded the parts originally written by you (in green), Gemini generated (in yellow), and DeepL translated (in blue). I kept the part you added/edited to the Gemini generated text in white and added them to the other two.
Everything in yellow is the AI generating what it has determined what you want to say -- not what you actually said in your writing. It isn't your writing. It's the AI's. The words it chooses to generate and the order it chooses to do so also raises a red flag for AI detectors.
As I stated earlier, you didn't write "The narrative maintains a sensible level of suspense through well-placed mysteries" in Polish. You wrote "niewiadome w plocie trzymają w sensownym stopniu w napięciu", which is "the unknowns in the plot keep you in suspense to a reasonable degree" in English. While they express similar concepts, one is out of place and not in your writing voice, at all.
I also mentioned that:
"While characters face serious problems... simplified and superficial" feels like forced writing, not natural.
The entirety Illogical Plot Decisions point sounds forced, as well.
The translated version "the characters have serious problems, but they are very shallow" reads more natural (though does lack clarity, but so did the original).
The translated Illogical Plot point also reads more natural.
Funny how you requested Gemini to make it shorter, but it is the longest of the three (100 characters longer than DeepL).
-----
I would recommend writing your reviews - fully and with proper Polish grammar, structure, composition - and then have a translation app properly translate your review into English. That way it is your writing, it is your review and it won't get flagged by any AI detector, if one is used to check reviews. Also, the translator's version will help you better strengthen your skills in English writing.
