New 'Vibe Coded' AI Translation Tool Splits the Video Game Preservation Community 43
An anonymous reader quotes a report from Ars Technica: Since Andrej Karpathy coined the term "vibe coding" just over a year ago, we've seen a rapid increase in both the capabilities and popularity of using AI models to throw together quick programming projects with less human time and effort than ever before. One such vibe-coded project, Gaming Alexandria Researcher, launched over the weekend as what coder Dustin Hubbard called an effort to help organize the hundreds of scanned Japanese gaming magazines he's helped maintain at clearinghouse Gaming Alexandria over the years, alongside machine translations of their OCR text.
A day after that project went public, though, Hubbard was issuing an apology to many members of the Gaming Alexandria community who loudly objected to the use of Patreon funds for an error-prone AI-powered translation effort. The hubbub highlights just how controversial AI tools remain for many online communities, even as many see them as ways to maximize limited funds and man-hours. "I sincerely apologize," Hubbard wrote in his apology post. "My entire preservation philosophy has been to get people access to things we've never had access to before. I felt this project was a good step towards that, but I should have taken more into consideration the issues with AI." "I'm very, very disappointed to see [Gaming Alexandria], one of the foremost organizations for preserving game history, promoting the use of AI translation and using Patreon funds to pay for AI licenses," game designer and Legend of Zelda historian Max Nichols wrote in a post on Bluesky over the weekend. "I have cancelled my Patreon membership and will no longer promote the organization."
Nichols later deleted his original message (archived here), saying he was "uncomfortable with the scale of reposts and anger" it had generated in the community. However, he maintained his core criticism: that Gemini-generated translations inevitably introduce inaccuracies that make them unreliable for scholarly use.
In a follow-up, he also objected to Patreon funds being used to pay for AI tools that produce what he called "untrustworthy" translations, arguing they distort history and are not valid sources for research. "... It's worthless and destructive: these translations are like looking at history through a clownhouse mirror," he added.
A day after that project went public, though, Hubbard was issuing an apology to many members of the Gaming Alexandria community who loudly objected to the use of Patreon funds for an error-prone AI-powered translation effort. The hubbub highlights just how controversial AI tools remain for many online communities, even as many see them as ways to maximize limited funds and man-hours. "I sincerely apologize," Hubbard wrote in his apology post. "My entire preservation philosophy has been to get people access to things we've never had access to before. I felt this project was a good step towards that, but I should have taken more into consideration the issues with AI." "I'm very, very disappointed to see [Gaming Alexandria], one of the foremost organizations for preserving game history, promoting the use of AI translation and using Patreon funds to pay for AI licenses," game designer and Legend of Zelda historian Max Nichols wrote in a post on Bluesky over the weekend. "I have cancelled my Patreon membership and will no longer promote the organization."
Nichols later deleted his original message (archived here), saying he was "uncomfortable with the scale of reposts and anger" it had generated in the community. However, he maintained his core criticism: that Gemini-generated translations inevitably introduce inaccuracies that make them unreliable for scholarly use.
In a follow-up, he also objected to Patreon funds being used to pay for AI tools that produce what he called "untrustworthy" translations, arguing they distort history and are not valid sources for research. "... It's worthless and destructive: these translations are like looking at history through a clownhouse mirror," he added.
It's the worst it'll be (Score:5, Interesting)
Geez, what a silly over reaction. AI translation is freaking amazing, and today is the worst it will be. Just put a disclaimer that the translation is the best available at this time and will be improved over time. The "I've canceled my subscription ire" is ai-phobia.
Oh là là, quelle réaction excessive et idiote. La traduction par IA est carrément incroyable, et aujourd'hui est le pire niveau qu'elle atteindra. Il suffit de mettre un avertissement disant que la traduction est la meilleure disponible actuellement et qu'elle s'améliorera avec le temps. Cette colère du style "j'ai annulé mon abonnement", c'est de l'IA-phobie.
Vaya, qué reacción exagerada tan tonta. La traducción por IA es malditamente increíble, y hoy es lo peor que va a estar. Solo pon un aviso de que la traducción es la mejor disponible en este momento y que mejorará con el tiempo. Esa ira de "he cancelado mi suscripción" es simple IA-fobia.
Meine Güte, was für eine alberne Überreaktion. KI-Übersetzung ist wahnsinnig toll, und heute ist sie so schlecht, wie sie nie wieder sein wird. Setz einfach einen Disclaimer drunter, dass dies die aktuell beste verfügbare Übersetzung ist und sie mit der Zeit verbessert wird. Dieser Zorn à la Ich habe mein Abo gekündigt‘ ist reine KI-Phobie.
(No Patreon Funds Were Spent Obtaining These Translations)
Re:It's the worst it'll be (Score:4, Insightful)
They couldn't give a fuck about the accuracy.
If you post anything using AI in any kind of open forum, a very vocal minority is going to accuse you of destroying the environment and the economies of the world.
The unemployed have opinions.
Re: (Score:1)
Don't apply high level advanced knowledge where there is none to be found. The irrational hate for AI is for a variety of extensive reasons. In the gaming world it's almost universally "OMG IT'S CREATIVELY BANKRUPT!" WON'T SOMEONE THINK OF THE POOR JOBS!
Re: (Score:1)
And they are not doing it because they think of the environment. You could set off you whole month of using AI by watching one fewer Netflix movie. Maybe even more, as AI is becoming increasingly more efficient. But that's not their point. They don't have a reason to hate AI, but they hate AI and are looking for a reason.
Re: (Score:2)
Watching netflix movies is overall better for the environment than the alternative - physical media being manufactured, shipped all over the world and ultimately becoming waste plastic, people driving to see a movie in a theatre etc. Something like netflix doesn't create much in the way of new infrastructure, it uses infrastructure that already exists and only contributes towards earlier replacement/upgrades.
Re: (Score:2)
I wasn't really arguing about Netflix being that bad. The point is, that the energy/use ratio of Netflix gives you a scale that the energy/use scale of AI is not a relevant problem. Google quantifies a query to their high-end model as 9 seconds of TV. Not streaming, just the device running. Unfortunately they didn't specify what TV exactly, but I guess we know the order of magnitude and if your TV is small you may be allowed to watch 18 seconds before the AI query is cheaper.
Re: (Score:2)
Exactly this.
The options are an imperfect ai translation, a most likely much worse translation using older translation software, or no translation at all.
Also note that those complaining about the ai translation are not offering their services to provide a manual translation. Most likely they can't, which also means they're in no position to judge the quality of the translation anyway.
Re: (Score:1)
You can't even see the irony, can you? What a fucking snowflake.
Re: (Score:1)
Sure, it's a trivial checkbox to post anonymously.
It's less trivial to comment on 10 of my comments within a 24 hour period with the same linguistic style, all utilizing the same basic insult and format, dumbshit. And yes, that's stalking.
Re: (Score:2)
Yes there is definitely a segment of the internet that melts down if they even see the word AI in any context. Makes it difficult to have any sort of objective discussion.
Personally, I am more concerned about the vibe-coded aspect. It's a tool like any other tool. It can be useful, or it can be used poorly. In particular I feel it's critical any "vibed" code be looked over by a developer who can read and understand it and verify it is operating properly, particularly with an eye for user security or privacy
Re: (Score:2)
Yes there is definitely a segment of the internet that melts down if they even see the word AI in any context. Makes it difficult to have any sort of objective discussion.
Sure the fuck does- and I've got plenty of negative opinions about it that I'd love to be able to have in good faith.
Personally, I am more concerned about the vibe-coded aspect. It's a tool like any other tool. It can be useful, or it can be used poorly. In particular I feel it's critical any "vibed" code be looked over by a developer who can read and understand it and verify it is operating properly, particularly with an eye for user security or privacy issues. Anyone who uses "vibe-coded" software that has not undergone such verification (even if only by its own author) is loading a hand grenade onto their own device.
I too share that concern. But at the same time- every programmer alive has produced shitty ass code.
I wasn't born an professional software engineer with 20 years under my belt.
And that's BEFORE you get into the possibilities of compromised AI agents having been instructed to embed malware into "vibed" code.
Ya. It's so highly fucking problematic.
But what the hell do you do about it?
Because shitty amateurs are arguably just as dangerous with their package inclusion.
I just find vibe coding a.... thing that makes shitty
Re: It's the worst it'll be (Score:2, Interesting)
Will it get better though?
I mean, AI translation tools will get better, but will the translations produced by current tools get better? Once material has been badly translated today, will someone actually use new tools in 2, 5, 10 years to update the "good enough" translations?
History tells us no. The risk in using AI tools for such a project is to create bad translations (and considering that AI doesn't do a good job today when translating technical or specialized subjects, they'll be bad) that will be nea
Re: (Score:3)
As long as they have the original text, it's only a matter of telling the future AI "hey, improve these translations by redoing them" and you'll have the new stuff in a few minutes. If at that time you still cared enough to complain, you can care enough to retranslate it.
The real question is, is it better in the meantime to have a possibly poor translation, or not have it at all. To me it's pretty obvious. Nothing's stopping you from ignoring what you think is a bad translation. Worst case you just learn to
Re: It's the worst it'll be (Score:3)
"It's only a matter of" may actually be enough of a hurdle if there's a "good enough" translation already, since ai translation isn't free, especially on that scale.
And as someone else said below, translation is hard, especially that kind, since video game magazines are full of obscure references, made up words, and sentences that, in a video game context, don't mean what the ai thinks. Ai translated video games and mangas is dogshit, for now.
I'm using ai translation tools, it's good for general stuff, but
Re: It's the worst it'll be (Score:2)
I'm sorry what is this princess and the pea bullshit.
If a Japanese as a second language volunteer translated those magazines why would you not display them, with some little attribution or note explaining they were done by community members or volunteers and not reviewed for accuracy by professionals, or whatever the appropriate standard is for ... a collection of historical video game magazines? Don't translate them unless it's the best possible because it will only be done once, what is that shit.
Re: (Score:2)
Once material has been badly translated today, will someone actually use new tools in 2, 5, 10 years to update the "good enough" translations?
Yes. As long as there are people with an interest in doing so.
The software will get better. The cost of processing will get lower. Some hobbyist will take it upon themselves to run the cheaply available tool on their own dime, or a researcher will contact the company and ask them to donate (tax write-off/marketing opportunity) tool time to their project.
This has been the way of things thus far and there is no indication that this will not be the case in the future.
Re: (Score:2)
Geez, what a silly over reaction...
Agreed. This recent article seems to explain the current technology and business implications well. [nytimes.com]
Re: (Score:3)
The library of Alexandria wasn't burned because people didn't like it, it was first party burned because of a special military operation, then after the occupation it fell into decay because knowledge was judged unnecessary for the occupied, and finally razed because its books were too woke for some fundamentalist dimwit with a grudge against ideas he didn't like. So, no, not the same at all.
Perhaps you're confusing the history of the Alexandria library with the one of Herostratus and the Temple of Artemis
What is this (Score:2)
I am unkomfortable with (Score:2, Funny)
People using software with built in spell chequers to write. I hereby will not be supporting anyone who posts anything using any fancy digital tool to right.
How dare people use tools to do work.
How dare they use funds as they see fit to achieve exactly what I paid for. It's unacceptable! I demand that they do their job the way I expect them too. Next time I sea any of this spell chequed bullshit I will be taking my $10 contribution elsewhere.
Regards
Karen.
Re: (Score:2)
If I pay someone to apply their editor skills to my work, and all they do is run the same spellchecker I could have run myself, then I'm going to be disappointed.
If millions of people are going to read all the material, it may make sense to 'bake' the result of an LLM so we don't have to re-run a lot. In practice, this material is not in that boat, relatively few people are going to be reading relatively small portions of the material, so they can LLM for themselves if they really wanted. So it's wasteful
Re: I am unkomfortable with (Score:2)
Umm. You would have the server generate a translation on demand and store the result, then date it for review or regeneration on access after some later date. Automatically compare the past version to the new one from a more recent model and flag major differences for human review. That's not even half a day of work, and most of that would be finishing the design I just spitballed, white board time, process design, not coding.
What's the point in making people redundantly do it themselves, it's obviously les
Re: (Score:2)
If I pay someone to apply their editor skills to my work
Let me stop you right there. Is that what you paid for? Is there any evidence anywhere that people on Patreon were paying specifically for human translation? If so you have a point, if not your post is incredibly silly and would probably benefit greatly from an AI re-write.
Re: (Score:2)
Given that these folks withdrew their funding upon finding out how it was being used, it's self evident that these people at *least* did not want to be paying for machine translation. Maybe they expected just acquiring and scanning materials or human translation, but machine translation actively drove them off. We don't have to speculate on whether these folks minded their money to go towards LLM translation, they said very plainly.
Re: (Score:2)
And that's the people I'm criticising. To be clear not just criticising, I'm insulting their intelligence, I'm calling them luddites, I'm saying these are the kinds of people who would get upset that a builder builds a house using a power drill rather than a good ol' fashioned screwdriver.
Now maybe they have a deep meaningful reason they want to have that opinion, in that case I'm going to insult their intelligence for not making an informed purchasing decision. When you care deeply about this topic the onu
Stop being stubborn (Score:4, Insightful)
Did you have a look at the "boycott projects that use AI" software? It is growing and growing. Because coders know how to use efficient tools. If you want to refuse anything that touches AI, you need to become a hermit. You're fighting a losing fight instead of embracing what great tools you get for free. You don't need to use them, but at least do not stand in the way of others.
Re:Stop being stubborn (Score:4, Insightful)
Lumping into "tools that you get for free" and building any kind of dependence on them is a TERRIBLE idea.
Those AI companies are TRILLIONS in the hole. Literally trillions. And they're going to want that back.
And when the true cost of a) training the AI, b) operating it, c) generating profit enough to d) pay back the investors is realised, those dependencies are going to tank any project reliant on them.
As it is they're having to ask for donations to run the AI to do this, what do you think's going to happen when that cost is 10, 20, 50, 100 times higher than today, but you're utterly reliant on it and all the good people who WERE on board before have fled because they were just ousted by AI and don't feel like coming back?
There's using a tool, and then forcing use of an unsuitable tool that in the future is not going to be free.
It's like the Bitkeeper / Git debacle, but times by several thousand.
If we all just lumped into every single programming paradigm as if it's the one true answer, and forgo everything that came before... we'd be in real trouble. We'd have been in real trouble a LONG time ago.
But at least those tools / languages / paradigms didn't come with a dependency on paying back their trillion-dollar price-tag.
Re: (Score:2)
So what? Let the AI companies die, the models you downloaded for free stay with you.
The worst that can happen is that you won't get more advanced models in the future. Yeah, then we just stick with the current status of quite helpful models and may need to fine-tune them to new programming languages ourself. Have a look at huggingface, people do it all the time. It is costly to train the base model, but fine-tuning with new data can be done on a couple of 3090s.
Re: (Score:2)
Models become out of date very quickly. Haven't you noticed? There's even a conversation about "if we were to only train on, say, data up to 1900, would AI be able to derive Einstein's theories" (the answer is "hell no", by the way).
The models are huge, proprietary, and have already "stopped learning" by the time you get them. They were trained on STUPENDOUS amounts of illegal data that, without a company willing to gather, process and take liability for, in the future you will find hard to ever gather a
What did we learn.. nothing (Score:5, Insightful)
I have worked with game designers, have been a translator, and am a developer. I have been a professional translator and had a friend who only did game translation. Yes I know designers hate AI and I understand it. I translated patents, corporate docs, scientific papers, manga, etc. Translation business has probably gone poof by now due to AI. But.. manga translation is actually really hard. Not only did the margins shrink to impossible due to lack of interest / budget for quality, it is also like translating film subtitles and sometimes even required historical research.
These days, I use Claude to check my Japanese before sending and sometimes to figure out obtuse emails sometimes. My written Japanese has improved. Claude is really good but I wouldn't use it for a final production film subtitle, game, or manga. For this purpose though, it would be amazingly fantastic as the sheer volume and the purpose of research means nobody overseas would be able to ever grasp a fraction of it or translate it all manually. As noted in the Ars Technica article: “Famitsu alone is over 1,900 issues, each with [a hundred-plus] pages,” journalist and author Felipe Pepe noted. “That’s one magazine from one country. [Human translation] would be ideal, but it’s impossible.”
I skimmed the GPL'd code, after 1500 lines got tired and got Claude to quickly scan it. It sounds like a really cool scanned document viewer with side translation viewer, but is not a translator at all. The guy in TFA apparently used some of the project's funds for translation of the scanned magazines which he mentioned were like $1 per issue which honestly, it is like 1000x cheaper than a human. No human could ever do it unless it was their lifelong hobby maybe. But, there must have been a lot of issues so I guess he racked up some charges and AI haters gunned him down, apparently.
If you really were learning Japanese (like I did long ago) you might try all kinds of scanning, OCR, online and offline dictionaries, and still not understand everything due to made-up fantasy words. You might have to corral some Japanese gamers to answer questions you cannot figure out. So this is not as good as having a Japanese otaku next to you but still pretty great and you could improve the translations too, couldn't you?
It looks like Nichols who hates anything AI deleted his post, and the person who spent tokens to translate is paying back to the project with his own money. And he made a GPL'd viewer which sounds nice. So it sounds like in the end, we have butt-hurt and afraid people, one honestly kind-hearted guy, some new OSS software, a free archive at archive.org, and a lot of translations this guy paid for. Hopefully other people might contribute translations either automated or not, so that people do not *at home* each pay for tokens to just translate their own copies of magazines that could have been just translated once and posted on the archive. I have no idea about copyright though I think one was from 1992. Nobody mentioned anything about that and if it is for historical research it sounds like fair use..
Why pay? (Score:2)
The worst part here, perhaps the only objectionable part really, is the use of community funds to pay for translation. You can download translation models and run them locally, and everyone involved is a gamer so they have a GPU to run them on. If AI translation is the only translation which will occur due to lack of funding for human translation, then nobody really has a right to bitch, unless you're wasting money on it... which is what happened.
Re: (Score:3)
Well, 'gamers' that are fixated on 20th century games, which does not suggest needing a vaguely capable GPU.
However, one could also connect the dots that an audience interested in the meticulous work of 20th century sprite based gaming that required the humans to do every last pixel might be the least receptive audience for 'AI away the human effort'.
Re: (Score:2)
Don't worry, I'm not one of his sponsors, so I get to look at all of this from a haughty remove.
Making your sponsors happy is how you get more money. I guess the guy is free to find new sponsors.
vibe code some Japanese lessons (Score:2)
You'd be better off learning Japanese than consuming AI slop.