As soon as I saw the title of this thread this morning, I knew this was probably going to be a hot mess…
Anyway, regardless of the usefulness of GenAI in language learning, we can probably all agree that AI has unleashed an absolute plague unto the world, a feature know as…
…Youtube auto-dubbing
NOOOOOOOooooooooooo…
While I’m at it, when I’m reading, the biggest issues I currently run into are identifying correct readings: 連濁, which number system to use for measure words, and when to use 音読み／訓読み.
Human native speakers will tell you that X and not Y is “natural.” Even in Japanese, you’ll see the word 自然. But ChatGPT doesn’t have intuition. It doesn’t know what is “natural”—or, conversely, what “natural” is.
Bunpro is one of the few places where I can read a sentence and feel confident that humans who grew up in Japan have vetted readings.
Actually the worst
But I think it falls in under the umbrella of “AI being crammed into spaces in an annoying way that people didn’t necessarily ask for” that makes this topic so salient. People are getting AI fatigue, not least with the latest “meta muse” feature that VERY quickly got rolled back
At least let us disable features that we don’t want to use!
I wanted to cry after seeing “AI friends” in LINE and HelloTalk
one time i scrolled upon an instagram reel and the auto ai translation sounded like what a radiowave would sound like 
it was pure static
Meta Muse was a terrible idea, that’s a perfect example of a terrible way to use AI. Still not sure who thought it was a good idea
yeah and dont bury the settings either… design is getting so hostile these days
it reminds me of when insta first added their ai search thingy so i talked to it to ask it how to disable it and it hallucinated instructions three times before i found out it couldnt actually be disabled 
I took a class on UI and it’s actually pretty gross how much you can get people to do with the way you word and lay things out. That’s why I think it’s really important to be very in the open about AI use and where and when it’s being used.
is that like “nudging”? weirdly learned abt that in philosophy class lol
Probably, I’m not the most familiar with nudging but it sounds like something that would describe it XD
nudging i learned about in the context of like how flashy cereals are displayed at kids eye height etc and stuff like that and how stores take advantage of decision fatigue to get you to buy stuff
wow i do not type like i took philosophy this is embarrassing LMAO
I was literally going to bring this up as an example but I didn’t wanna be wrong. I feel like I just had my mind read that’s wild
I just think it’s important to differentiate machine learning broadly from LLMs, and AI is a word heavily associated with LLMs (and stable diffusion models, which are based on similar principles).
Right now it’s good business for people to lump those all into the same bucket, but companies that specialize in other types of machine learning will soon be desperate to distance themselves from LLMs.
I personally don’t see the need for GenAI in Bunpro. While I’m sure people have various reasons for using Bunpro, at it’s core, it is a tool that provides high quality content written by native speakers, with the added perk of an SRS system. If you do a competitive analysis of what other tools are commonly used by Japanese language learners, the top tools are likely Anki, Wanikani, and Duolingo. Each of these has very obvious drawbacks (which I’m not going to get into because that’s a whole other topic). At least for me, Bunpro provides so much more value than these other apps because it has so much native written material and makes it easy to access that material in a customizable way. Adding in GenAI content would make me question the reliability of that material, and I think would actually make it stand out less in a crowd of competitors.
Another thing I don’t think anyone has mentioned yet is: why should the Bunpro team prioritize this over other features or general site maintenance? The Bunpro team is relatively very small. The past few years, they’ve been releasing a ton of new features such as Topics and Mock JLPT tests. Those features are great, don’t get me wrong, but you can already see how by them prioritizing those features, other content and updates have been pushed to the side. N1, A1, and A2 translations still haven’t been completely verified by staff, even though they were released close to two years ago. The A decks still aren’t accessible without a direct link. The Bunpro Summary sections and Hints still need to be updated for most levels. There are a ton of great quality of life improvements that have been suggested in the Feedback - Suggested Improvements/Feature Request thread. Randomly adding in GenAI tools just because it’s the popular thing right now would further take away resources from core areas of the site that still need to be worked on.
As others have mentioned, old school AI/ML is a different topic, but for GenAI at least I don’t see how integrating it into Bunpro would benefit me as a user. There are already readily available tools out there that people can use if they want to. I don’t need Bunpro to be an all encompassing tool that does everything, especially if it comes at the cost of quality and the team’s ability to maintain its existing content and features.
The reality is that, right now, there is not much that Bunpro offers that you can’t get elsewhere, or even for free.
Bunpro’s value comes from ease of use. The convenience that comes from having everything in one place with a slick interface and stat tracking.
All of which can be done, trivially, by AI. Bunpro is an easy bolt-on for anyone with a successful AI based language learning product. And when customers have to choose between Bunpro and Bunpro + other successful product, they’ll choose the combo pack every time, unless they’re a Luddite of some flavor.
I don’t really know what you expect for proof. We’re speculating about the future here. I’m sure the horse and buggy people were demanding proof that noisy, road reliant cars were going to displace quiet, road independent horses. But by the time it arrives, though, it’s too late. You’ve lost.
This reminds me of my friend who was like, I don’t wanna buy a car. The second I buy a car I know im gonna be using it just to cross the street.
And I laughed, because it was funny, but also because he was right.
Sometimes there’s also a valuable reason not to take shortcuts, just because they’re there.
AI might be the future but I think people have brought up plenty of great reasons why it doesn’t need to be or shouldn’t be Bunpros future