It would be great to see a use-case.
Otherwise - I see a little value to do something for the sake of having “AI” in it.
3 Likes
Please don’t ruin Bunpro by bringing AI into it.
14 Likes
I think we could convince @bunnypro to be the face of BunCoin. I wonder how many carrots we would need…
5 Likes
You could twist my front paw with that magic combination of B-points and carrots 
4 Likes
For personal use, AI has been beneficial to me when it comes to Japanese, but I can see why a pro website would be reluctant to implement it.
Bunpro places value on it’s accuracy and authenticity and if AI get’s something wrong, they’ll take the hit for it. Look at what it’s done to Duolingo.
AI has been a great tool for me personally and explains and does things that go way beyond what a human made program can do, but it also does a lot of really strange and unpredictable things. That’s why it’s better utilized to fill in the gaps.
6 Likes
I see what you did there 
2 Likes
Let me provide an alternative view.
Bunpro took off, as one commenter said, because there was a gap in the market for people explaining Japanese grammar as well as Bunpro did, especially all in one place.
Extend that logic.
If there is a good way to use AI to help language learners someone will find it. And the businesses (and Bunpro is ultimately a business) that fail to adopt this beneficial approach will lose customers to those that do.
AI is here. Bunpro can adapt, or it can lose market share. Those are the only options on the table.
So I would suggest that the question is not ‘should Bunpro adopt AI?’, but rather ‘when and how should Bunpro adopt AI?’.
Just my two cents. (And yes, it absolutely must be locally hosted and owned outright by Bunpro. You cannot allow the quality of your product to vary based on the periodic updates of some disinterested third party.)
2 Likes
@SilentRequiem You make the incorrect assumption that AI is an “adapt or lose market share” type technology with no substantial backing to the claim. Bunpro (like textbooks) is not a technology driven product: Bunpro could’ve existed in 2011. Good quality language products exist for their quality and trust in their product.
Bunpro’s unique selling point is the quality of their content put in high quality package - not any fancy SRS mechanism, novel UI or marvellously new ideas. Stuffing AI into a product like this just to “keep up” spoils that uniqueness.
If Bunpro doesn’t catch onto AI it’s not going to die out like Blackberry did in the wake of the iPhone. Textbooks haven’t died out in the wake of smartphones.
I’d like Bunpro to thrive, and there’s a lot more compelling business reasons to not force AI than there are to try and “innovate” in a product that doesn’t need innovation. Bunpro risks dying much more due to needless trend chasing just to not get “left behind.”
And suppose Bunpro does die out due to some revolutionary idea? Well I’m sure whatever replaces it will be better.
25 Likes
This is exactly why we shouldn’t integrate AI though. It’s divisive and the site doesn’t desperately need it.
6 Likes
i think an ai chatbot would help people have conversations. i use chatgpt to help me form complex sentences and i find it very useful.
ai can be good if you use it the right way
but i dont think the site should be completely ai ridden. maybe a tab in the practice section to practice with ai. i would definitely hate it if bunpro adapted ai in everything
2 Likes
I agree here sort of as I detailed in an earlier post. There is a space where AI could help bridge the gap between doing SRS drills, and actual writing output. Even if executed well, that’s low on my list of desired features.
3 Likes
The intention was more so to try and help combat against the growing trend of AI = BAD regardless of if it helps anyone.
I’m honestly a bit surprised by how much there is this push against AI in large part for the reason of well I just don’t like it.
I’ve seen good criticisms like reliance on other companies, price, reliability concerns, but also a good handful of people offering no value to the discussion and just saying AI is horrible and you’re terrible if you it.
Even if you were to get specific and say Machine learning many people still associate that under the idea of “Evil AI”. Even the Bunpro team has at least tried it because they’ve seen some potential in it. It might not be perfect today but as I’m sure most people have seen there’s a lot of money being poured into the development of AI. There is a good chance that it will get to a point of quality that Bunpro would even deem good.
Unfortunately I still think at that point people will push back, even if it was free and guaranteed to help them learn 2x as fast. I want to see features like personalized lesson plans, review times, help with writing and reading. I think the ability to ask and get an explanation back in Japanese using words and vocabulary I know would be great.
With all of this I can also recognize and respect where other people are coming from but I’m seeing a growing amount of people not even reading or trying to understand and just taking the stance of you’re wrong I’m right.
I can say that as a lot of GenAI is today it might be hard to implement and not be worried about the quality of the results. I do believe that ML is good enough to see larger use in improving and personalizing recommendations and reviews for users. I also think that’s it’s just a matter of time before GenAI gets to the point of being good enough for the Bunpro team to be comfortable to consider adding features with it and I’m excited to see what come!
2 Likes
i agree with you, most people just see “AI” and immediately think “bad!” without any reasoning or background information on the source. i do use ai like chatgpt to help me learn and with other stuff not related to japanese like researching. but i also think generative ai, like ai “art” generators, are horrible because they take people’s original artwork, and somehow makes it worse. its even worse when ai “artists” post their “art”.
this has became an ai rant whoops.
but i do believe bunpro should integrate ai in some way that does not affect the whole website.
2 Likes
it would be great if Bunpro documented their APIs and released an MCP for easier agent interaction. I built some flows to help manage vocabulary and figured out the endpoints needed through trial and error, but the agent knowing where it needs to go to find data would be really helpful.
2 Likes
As others have mentioned, I’ve used Gemini and ChatGPT to spell check and correct my free writing exercises and grammar drills. It’s the only alternative at hand since I don’t have a tutor and don’t feel like using those ‘post sentences, get reviewed’ websites.
In fact, one thing I do is copy paste the grammar point’s titles from Bunpro plus their English translation (so that the LLM can tell what usage I’m testing myself with) and write my own sentences, usually compounding other structures to string together a slightly longer sentence.
Been doing it for a while and in my experience, it kinda sucks. It’s better than nothing, sure, but I’ve gotten to a level where being skeptical makes perfect sense.
It sometimes mixes up usages. It corrects me on what sounds “natural” or not, even though it’s not really prepared to do that. Since I’m not going to write a full short story to provide it context for every single grammar point, it assumes things in odd ways, over correcting.
Even with a base prompt like “do not rewrite my whole sentence” or “point out the mistakes but don’t tell me why” it’s been very frustrating.
The grammar break downs sometimes are a more natural rewording from some textbook, that may or may not actually apply. When pressed, or when asked to reformulate it in X way, it hallucinates. It feels like I’ve learned a thing or two from these interactions, but being on guard at all times is tiresome, and what’s worse, at different stages, I could not really tell it was pure BS. One time I even got it to quote a fake paragraph and fake page number from the Dictionary of Japanese Grammar. I didn’t trust it then, and was never going to trust it afterward that. Asking it to grade my level is also another lost battle.
I’ve spent rought 1/3 of the time arguing with it instead of actually writing anything.
What I found it useful for is having it pick up random grammar points I should know by now, a setting or topic, or feed it a list of “leeches” (kanji I just can’t remember) when I’m too lazy to come up with things on my own. I use its prompt to write things on my own.
I would not trust it to compose a text and if I were to read AI generated stuff, then I’d like to at least know it’s AI generated. Other platforms have done this, and filled their entire reading sections with this kind of articles. Even an intermediate student like myself can tell them apart.
8 Likes
Could you possibly message me the prompts you are using? I’ve been working with making generic prompts to create a tutor of sorts using different models and I’d love to test as many directions as possible. I might even be able to get something that can help you with your own personal usage!
Again I feel like all this discussion comes back to the fact that people are free to use AI to supplement their studying in their own way so I don’t see why it needs to be added when adding it would make so many people upset without adding significant value. Regardless of whether people are wrong for disliking AI, the fact is user sentiment is only going to become more divisive as AI presents more risks, even if those risks wouldn’t even factor into a bunpro use case scenario. It just doesn’t seem worth the alienation for minimal gains when people who are going to use AI are already doing so
8 Likes
My problem with ai correcting my sentences, is that it corrects every bit of it, and makes me feel bad, and it cannot say how that mistake would sound to a native, would it sound bad? Wouldn’t native even notice? Grammar is okey, but then it tries to correct my word choice and again I cannot say if it’s neccecery or it just decided that “this looks better” like to day it corrected my 気がする to と思う and said not to use しても、それとも at the same time (not piling them up, which is obviously right not to do that, but in a casual speech and fast writing it’s okey I think, I cannot say).
And every time it makes a correction I just cannot believe it, because I know it doesn’t know nuances, and all this leaves me in some strange state of my mind that I don’t like. It’s kind of unsustainable for me, and I imagine for other people like me.
Grammarly on the other hand is quite chill, but still I would prefer a real human until my language becomes perfect
5 Likes
my chatgpt does everything i say it to and predicts what i mean even without me telling it.
im pretty sure you gotta change it in the personality menu, the screenshot is what i set mine to 
and for custom instructions, i mentioned that i am studying for N5 so it uses basic vocab and makes it explain everything it writes
1 Like
It told me something I said was super weird and then my tutor was like huh no thats fine
It’s risk averse when it comes to language so it will default to the most common way of speaking it can find on the internet, and I think that harms user improvisation and experimentation
10 Likes