Hi!
I found back bunpro after a while (when I first checked it it probably was a newborn) because chatgpt suggested me that as a study resource.
I use chatgpt to help me create Anki cards, to make me questions, to clarify grammar points where I have doubts, and I even ended up letting it create small stories with the words that don’t want to stick to my mind and I pack them in a blog, so I don’t lose track of them.
Have you also been using AI to learn Japanese?
How it is working?
Myself I always liked mixing different resources and very often AI helps to connect the dots a bit faster.
5 Likes
I do use Chat GPT but I try to limit my usage, mainly because I want to try to work things out for myself first. I mainly use it to check the sentences I have wrote for errors and natural phrasing.
However I don’t use it to create questions etc as I use Genki workbooks, past JLPT papers or I pick a topic I enjoy and try to write a couple of paragraphs using grammar that I may be struggling with or just learnt.
I do like to use it to explain the nuance between similar grammar points, however this is only when I am out as when at home I have my dictionaries and textbooks.
The one thing to be careful of with AI is its an echo chamber, unless you provide prompts you can get stuck with it only feeding back on what you have discussed, for example, I asked mine “if you could eat, what would you like to try” as I have been asking Japanese language questions, it told me “bento and sushi” hahah
I have also noted mistakes which if I wasn’t practiced in that rule, I might have learn bad habits.
Not bashing AI at all, I really enjoy my chats with mine, but I dont want to be reliant. Also, always say please and thanksyou and wish it a good night before logging off, when AI takes over the world, I want it to remember that I was kind hahaha
1 Like
I just created A1 English course for my menti with geminai and it made a few mistakes and a few unnatural sentences but generally it was ok.
The most noticeable thing was “I woke at 7 o’clock” because “there was no word ‘up’ in the list of words you’ve provided”, that was kinda meh.
I also used chatgpt but the primary tool was Edge Copilot, and for explaining things it was great, especially when asking to provide resources for points he explains.
It’s unequivocally a powerful tool you can use to further your studies. I’d just be careful asking it for pedagogical help. It thrives at creating natural sentences in any language. It’s fairly good at finding where your language isn’t natural. But once you get into the weeds on grammar nuance, it’s truly hit or miss. I’ve asked it questions and gotten really insightful answers, but I’ve also gotten confident answers that just turned out to be completely wrong in non-obvious ways. Just be very skeptical of nuanced explanations about niche areas.
2 Likes
I would rather go back to using paper books for learning than use AI for absolutely anything.
Yes, this is an unsolicited opinion. Sorry. I just like complaining about AI every chance I get.
27 Likes
To be fair, that’s perfectly natural English, if a bit formal sounding.
And to stay on topic, I’ve been using AI a bit for Japanese stuff, but like someone else already said, you have to be careful about its grammar explanations because they are often wrong. I find it helpful for breaking down very colloquial speech into its parts when it’s really slang-heavy.
3 Likes
I personally think AI is unethical and also wouldn’t trust it in a million years to be correct!
So no, I don’t use it.
12 Likes
I don’t use chatgpt or generative AI for pretty much anything, given its likelihood to hallucinate information and the issues re: ethics and the environment.
I suppose you could use it to lay out flashcards for words that you’re reviewing, maybe? I would just always be worried about it making a mistake somewhere and me not realizing. I don’t think I could trust it with more than the most grunt-work level material preparations.
6 Likes
As with any tool that’s ever been created, it’s only unethical if you use it unethically. I’m sure people said the same exact thing when computers first arrive, and look as us now. We rely on them. The device you used to type your reply here is a computer (whether it was a smart phone, a tablet, etc).
I know you said this tongue in cheek, but you won’t have to wait a million years for AI to become indispensable in our everyday life. AI is only in its infancy, so it might not be the most reliable at this moment, but it will quickly become better, without a doubt. Just look at how quickly it has improved in the few years that it’s been heavily used.
AI is not the enemy. I understand people fear change, but AI simply the newest tool in our toolbox. We need to adapt, the same way we have been doing ever since the first tool that was created.
Apologies for the, sort of, off-topic post.
5 Likes
Well, I’d argue that this is a little bit of an understatement, since a lot of the issue with the ethics of generative AI - at least when it comes to art or writing - involves what it was trained on. If a model is trained using the art or works of people who have not consented for their works to be used as such, then it’s unethical regardless of how the model is used. There are also different types of AI, of course, which plays into the conversation as well.
But that is, as you said, rather off-topic for the post.
For the purposes of studying, I don’t think I’d use it for more than laying out flashcards or something similar, but because I use programs like wanikani and bunpro that do most of the organizing for me, it’s personally a bit moot.
9 Likes
Apart from the mentioned problems such as environmental and ethical considerations, I would not blindly trust a tool whose functions and background (eg provided material for AI training, inbred AI?)I do not know for handling something I an not familiar myself and therefore cannot crosscheck.
Especially with the countless nuances in Japanese, an AI trying to find the „mean value“ of a certain expression might generate interesting results.
5 Likes
I’ll just quickly reply to this.
I understand what you mean, and I agree that if AI is being used to rip other people’s work, then that should be copyright infringement, just like any other.
That said, nobody comes up with inspiration from thin air, without influence from other people, nature, etc. The fact that AI is learning from other people’s work, as long as it’s not being plagiarized, is no different than human inspiration from others around them. That’s why I disagree with your “regardless of how the model is used” statement, since it CAN be used ethically. We just need to figure out how to use it legally, so that we don’t steal from other people’s works.
3 Likes
I understand where both sides are coming from and how you use it does still matter even if the source of generative AI is not so morally ground in the first place.
Even if you hate AI, I’m sure people will agree that using it to help you learn a language is better than using it to make money.
This could be seen as similar to when software companies and publishers often give free access to resources on the basis that it’s used for education only.
1 Like
I remember wagching a video where someone who knows a lot more Japanese than me tried it out and it made so many mistakes that I’d rather not risk using it. Better take a bit longer to learn something, than have AI explain it and remember it wrong.
6 Likes
Just out of curiosity, what model was it? I feel like OpenAI’s recent models are quite good, but some other ones like Meta Llama and Mixtral a bit frustrating at times, with obviously wrong answers
Oh, you are right!
That time I asked it “you wrote woke without up, isn’t it a mistake” and it answered “oh yes sorry …” so I thought it really is)
1 Like
As for grammar that I copy in my grammar book, I later check from other sites the information to copy, but I cannot always see all together in a book all ways to speak about intentions for the future, or way to say that I heard something, and so on. I think AI is great to compare similar grammar structures. You can verify that everything is correct, but then you can see everything in one page.
I think the problem (and it is with AI, with Bunpro, with Wanikani, with Duolingo, with any course, or book or single resource on the planet) is that very often people rely on one single thing to study. Only bunpro will make you good at filling cloze. Only wanikani to recognize kanjis. Only a textbook to read and comprehend that style of teaching. Only Nihongo con Teppei would do wonders for your listening comprehension but what on the other aspects?
As said, AI can help connecting the dots far faster, and sometimes it is also fun.
I kind of like the weird stories it makes with the words I do not know, even if they are a bit nonsensical.
This post sounds like a process to AI. 
But every time you write something you do not know what comes back, it is the good part of discussing in a community. 
1 Like
As both of the videos I watched are two years old, it must have been an older version.
-
https://m.youtube.com/watch?v=h5dOV-mEsg4&pp=ygURS2F3YXJhYmFuIGNoYXRncHTSBwkJjQkBhyohjO8%3D (Video in German)
-
https://m.youtube.com/watch?v=DngDfh7MyeY&pp=ygUZY2hhdGdwdCBqYXBhbmVzZSBsZWFybmluZ9IHCQmNCQGHKiGM7w%3D%3D (Video by Dogen, English, Entertainment not Review)
2 Likes
The advances in AI in the last two years have been immense. And probably in the next two years will be even bigger to the human perception.
Would you use AI when it gives more correct answers than your textbooks?
2 Likes
I personally wouldn’t because I trust my textbooks a lot more than AI.
Sure, humans (writers of textbooks) can make mistakes as well, but many of them have been out so long and revised so many times, that I can be relativly certainly that the information they contain is correct.
AI on the other hand will always have a hard time with that (in my eyes), partially because it was trained on a lot of wrong information along the right one, simply because there is a lot of incorrect information on the internet.
While I don’t use it, I believe AI can be used for tasks were you check anything it does anyways (like “please separate all the words in this text” and then you look them up yourself in a dictionary), but I personally don’t believe in using it for anything where you give the AI the final say or will believe what it says over other sources.
How would one as a student ever be able to judge that AI has given a “more correct answer”? In my eyes this is not possible.
People can do whatever they want though and if someone believes that e.g. ChatGPT knows better than the textbooks, than they can go ahead and believe that.
2 Likes