English translation for bunny black 3

xRoguex

Well-Known Member
Aug 2, 2021
1,096
1,389
287
but this only translates some of the text like the prologue fully and after that some parts is translated and other parts arent.

can you tell me how to upack them so i can pass everything with chat gpt to translate the parts that are still in jap.

while a proper human translation would be better i am sure a lot of people would be ok with just a gpt thanslation. since playing with OCD translation is a pain in the ass.
can help you probably. But it doesn't work with every game. I used it for getting pics and bgms.
 

millyrainbow

Member
May 11, 2018
191
136
86
Problems is the money origin
money, ai training, ai training cannibalization (which is actually pretty bad for some models), and also the ai having a bad habit of just making shit up, even when it knows exactly what the real answer is.
In many cases i would still be running it through google translate just to make certain the ai is actually telling the truth.

Though running it through an MTL and then having that pushed through ai as an interpreter might work, so long as the ai doesnt break and start converting single sentences into paragraphs, or feeding off previous entries to make duplicate entries, or any other number of things the ai can do because its borderline autistic.

I have a lot of experience working with ai, and sometimes i just really really want to beat it with a baseball bat because of how stupid it is.
 
  • Like
Reactions: lurisfantasy

Boarborn

Member
Apr 11, 2022
128
123
167
soon AI is going to do everything necessary to translate it
we should ask china, micrzoft, googi, elon... defeating a curse it's a big show of their capabilities :)
 

millyrainbow

Member
May 11, 2018
191
136
86
soon AI is going to do everything necessary to translate it
we should ask china, micrzoft, googi, elon... defeating a curse it's a big show of their capabilities :)
I wouldnt put too much faith into seeing that. The more complex the task, the more ai falls apart. AI can barely even interpret things, and thats when its not lying to you or straight up making shit up.
 

bllahbllub

Active Member
May 5, 2017
849
2,517
352
I wouldnt put too much faith into seeing that. The more complex the task, the more ai falls apart. AI can barely even interpret things, and thats when its not lying to you or straight up making shit up.
Yes and no.... mtl exist since the early 2000 and that was the only thing available.... everything from japan was mtl and horrible, that's why i can say ai can do a good "clean" translation without problem the proof reading is the problem...
 

millyrainbow

Member
May 11, 2018
191
136
86
Yes and no.... mtl exist since the early 2000 and that was the only thing available.... everything from japan was mtl and horrible, that's why i can say ai can do a good "clean" translation without problem the proof reading is the problem...
If we're just talking strictly taking the text and translating, while peeps do the rest of the work, then this is partly true; at least if we're talking about LLMs, which from what i understand, is the direction many seem to want to go. The largest problem with LLMs is they can lie or make stuff up even when prompted to not. If used alone its very unreliable, even with proof reading.

I am not sure how close or far we are from getting LLMs to be reliable. I personally dont think we're anywhere close to that though. Theres a lot about AI that has already peaked thanks to how its development is stuck on mimicry and its going to hold everything else back until we get past that.
 

Penitensary

Active Member
May 10, 2020
967
987
146
I wouldnt put too much faith into seeing that. The more complex the task, the more ai falls apart. AI can barely even interpret things, and thats when its not lying to you or straight up making shit up.
Pretty much.

See it like this: Every twenty years or so there's an attempt at commercial VR. Nintendo tried in the 90s and Google in the 2010s, so we can safely claim it wasn't due to a lack of funding that they never really caught on.

Like with every new thing, AI needs to fail horribly before silently getting some tweaks and tech updates in the background and giving them another go x years from now. It's part of the process: Hopefully next time they won't steal artist's work, that was basically what killed it in the eyes of most people that would consider using them. Currently, AI is at a level of 'Better Than Nothing', and better than nothing simply isn't good enough. The only uses i've mostly seen it used for are porn (which isn't setting the bar high) and those zero effort AI narrated recap videos on Youtube, so better luck ten years from now i suppose.
 

millyrainbow

Member
May 11, 2018
191
136
86
Pretty much.

See it like this: Every twenty years or so there's an attempt at commercial VR. Nintendo tried in the 90s and Google in the 2010s, so we can safely claim it wasn't due to a lack of funding that they never really caught on.

Like with every new thing, AI needs to fail horribly before silently getting some tweaks and tech updates in the background and giving them another go x years from now. It's part of the process: Hopefully next time they won't steal artist's work, that was basically what killed it in the eyes of most people that would consider using them. Currently, AI is at a level of 'Better Than Nothing', and better than nothing simply isn't good enough. The only uses i've mostly seen it used for are porn (which isn't setting the bar high) and those zero effort AI narrated recap videos on Youtube, so better luck ten years from now i suppose.
A lot of people always get confused with image ai. People think its all one thing, its not. Image recognition will continue to get better, its far from peak. This is not the AI though. Its just a basic pattern recognizer program that tries to replicate patterns. Thats it. The ai on the otherhand is just as dumb as it was 20 years ago. Its a basic interpreter that is incapable of understanding, and only capable mimicing and assuming. All it sees is what you feed it and tries to a: interpret what you said, and b: convert it into usable labels for the image recognition to do its job of pattern mimicry. Its that part A that is lacking.
This is what i mean when i say that AI as it is has already peaked, and it peaked a decade or two ago. The fact that we cannot get past the mimicry stage is what is holding back every tech that relies on it. And the breakthrough thats required to push us forwards is one thats going to be so massive that it will literally change the world over night. It will likely be even bigger than the microchip.
I do not forsee us seeing that breakthrough for another several hundred years though, but who knows.
 
  • Like
Reactions: lurisfantasy

lurisfantasy

Newbie
Apr 2, 2018
22
6
126
A lot of people always get confused with image ai. People think its all one thing, its not. Image recognition will continue to get better, its far from peak. This is not the AI though. Its just a basic pattern recognizer program that tries to replicate patterns. Thats it. The ai on the otherhand is just as dumb as it was 20 years ago. Its a basic interpreter that is incapable of understanding, and only capable mimicing and assuming. All it sees is what you feed it and tries to a: interpret what you said, and b: convert it into usable labels for the image recognition to do its job of pattern mimicry. Its that part A that is lacking.
This is what i mean when i say that AI as it is has already peaked, and it peaked a decade or two ago. The fact that we cannot get past the mimicry stage is what is holding back every tech that relies on it. And the breakthrough thats required to push us forwards is one thats going to be so massive that it will literally change the world over night. It will likely be even bigger than the microchip.
I do not forsee us seeing that breakthrough for another several hundred years though, but who knows.
"AI seems to mimic Satan, pretending to be God."
 

Penitensary

Active Member
May 10, 2020
967
987
146
Images or text is just the output, ultimately as unimportant as whether you want it to answer you in English or Swahili. But yes, we're still at the point where it is still too erratic to be a viable option.