Anki addon for interacting with OpenAI to learn Japanese

[LegenDX]

Hello, Bunpro community!

I have been using Bunpro for a long time, and it really helps me to learn Japanese.

Besides Bunpro, I used Anki for vocab.
Recently, I developed one Anki addon: Anki Quick AI, which is used in Anki to seamlessly interact with OpenAI.

For example, I am using it to generate a story containing all new words I learn today.
Here is the example video

The function of this addon is to gather card information with Anki browse query, and then send this information included in your customized prompts and get the response from OpenAI. I also support speech generation for the response.
The idea is that if we want OpenAI to generate something based on what we have learned in Anki, the card information may vary every day but the prompts may remain the same.
So “quick” means to configure prompts once, and then get responses from OpenAI with one click later.

Please feel free to provide feedback, suggestions, and report any issues you encounter. Your input is valuable in helping me refine and enhance the addon further.
I may be continuously working on improving and expanding the capabilities of this AI Addon. But as a Ph.D. student, I can only develop this in my spare time.
I have no prior experience in developing Anki plugins, and this is my first plugin. It may be buggy and there may be aspects that might not meet your expectations. I would be grateful for your understanding and forgiveness. Thank you!

Let’s embrace the power of Anki and OpenAI together!

You can find details in my GitHub repo
Here is the add-on page
The add-on code is: 547821970

[simias]

That looks really cool. I don’t know if I trust these AI enough to use something like that at this point, but it really gives an insight into what language learning could become. I think within the next few years we’ll get some tools that would have looked like science fiction a couple of years ago.

[username2]

@simias

I think within the next few years we’ll get some tools that would have looked like science fiction a couple of years ago.

Anki is already the tool that is still looking like science fiction for most people. For most people around me, learning language = read some textbooks, watch yt videos… ??? Some of them might use classical paper flashcards at best. What is spaced repetition? Screw it.
When I show them Anki their response is usually “ehhhh… too difficult”. Still, most of them impressed how I remember new words. If I should compare, I believe the same reaction was in medieval time to people using Arabic digits who could just calculate their stuff in an instant so quick, they were considered heretics at the beginning.
And it’s really difficult to explain this thing to people while still being in System 1 concepts. SRS is not shown in morning dramas or popular movies. So recently I gave up and just saying I’m putting a lot of efforts into my learnings. Not like it’s not true though, but most of people don’t do it as well, so this justifies things in their head without squeezing them out to System 2.

And for OP, I’m really excited with your approach! Totally looking forward to the way we can use OpenAI in language learning!

Which making me think: do we still using machines to learn language better, or rather it already machines start training organic-based neural networks to be useful for their industrial world? They follow-up our progress and feed us with reviews right at the good moment, they encourage us with good marks and punish with making to repeat failed reviews… Now they can create reference sentences like a proper teacher would do!

[davebp]

With seamless, instantaneous, perfect translation available, why will anyone bother to lean new languages?

— Dave

[kelth]

because there a no perfect translations. Language shapes cultures and ideologies. If you aren’t working directly with the language, you miss a little of the nuance, no matter how perfect the translation. This means you miss a bit of the ideology and culture behind the words as well.

-serious answer to what was probably a joke

[davebp]

I appreciate the serious answer. It wasn’t entirely a joke, but certainly tongue-in-cheek.

I fully agree.

[username2]

@kelth

Language shapes cultures and ideologies. If you aren’t working directly with the language, you miss a little of the nuance, no matter how perfect the translation. This means you miss a bit of the ideology and culture behind the words as well.

Let me add here a quote of Benjamin Lee Whorf that I love: “Language shapes the way we think, and determines what we can think about.” This means, literally speaking in different languages, we’re speaking slightly in different world’s coordinates. I’m three-lingual and I totally feel being a different person when I speak each of languages.

So personally, when I speak about language learning being done properly, I’d speak not about “learning foreign language”, but about leaning a new mother tongue. Don’t try to “translate” your thoughts to another language - this is just a hindrance. Think in your new language and speak in the way natives speak in it - to achieve your goals. It might be difficult from the beginning, but this is the way to freedom, not just fluency.

[LegenDX]

Both the answer that no perfect translations from @kelth and language shapes your mind from @username2 are good reasons.
There are actually many discussions about whether we still need to spend so much time learning a new language in the AI period from last year when ChatGPT was released. Suppose everybody wears some intelligent product like Vision Pro or Apple Watch that can provide instant translation, do we really need to learn a new language?

To be honest, I am pessimistic about learning languages in the future. The thing is that you need to spend so much time to be able to use it in daily life, and even more to read or watch native content. Also, you may quickly get worse if you have not used it for a long time.
But still, even though you spend such a lot of time, you would not be comparable to AI in language translation or polishing or anything else as a non-native speaker.
From my perspective, the input-output ratio of learning a new language will become lower and lower with the development of AI.

For most people whose native language is not English, maybe it is only a good idea to learn English because it is the most widely used.
But for other languages, learning it is just like hobbies or interest. Just like you can easily listen to the songs you like played by famous and skilled pianists, do you need to learn to play piano?
For me, I learn Japanese to watch anime, play Japanese games, and speak with Japanese in real life (though I have not met any Japanese until now). Mastering Japanese does not benefit my career at all, so I learn it only for my hobbies.
AI does not hurt my intention to learn Japanese. For example, previously when I have not started to learn Japanese, even though there are English or my native voices for some anime, I still watch anime with Japanese soundtracks and in other languages subtitles. My goal is not only the translation, so even though AI can provide instant translation (or for videos, it does not need to be instant at all), I will still learn Japanese.
On the contrary, if you learn a language for your career such as a translator, I think the future is not bright.

[Ishvalan]

That’s pretty amazing, I’ve only just started learning Japanese in the last 3 weeks but the amount of learning tools available is so great. I’ve been studying a 1000 most common words deck on Anki since learning hiragana and katakana. The deck has example sentences and native recordings but this tool seems like a great way to generate practice readings. I need to do some research on what other tools have been made for Anki.

[JCalandr]

I’m planning to develop some tools one day too … but for AR/VR Imagine you wear a headset all day to work, and have a small widget in the corner to learn your kanji/grammar/vocab, or even non-japanese stuffs. Like a small part of your everyday-space to learn anything.

Sadly, I’m a python developer … so it’s not suited to creating apps.

So I started learning Swift to be able to create tools like that for the Apple Vision Pro (imagine how good it could be to have a small bunpro widget on the side of your screen, so every time you have a 30sec break you can do 1 bunpro or anki rep <3)

First, I need to learn the language … then do an anki-like app, Text to Speech (the apple voice is quite smooth), a connection to Jisho.org, perhaps even GPT 4, … And I’m working on AI, so I could try to train a model with language learning in mind (kanji keys, pitch, linguistic books … instead of using the model for every type of question …)

[LegenDX]

Definitely agree that with Apple Vision Pro there will be many more creative and interesting ways to learn languages, which we have not thought of until now
I have no experience with Apple Vision Pro (though maybe nobody has), so I am a little curious if Apple Vision Pro can directly run iOS or iPad OS App, then we can directly run Anki in the corner or run it through a web browser in the corner.

Also, Anki is actually very powerful in that the card is in HTML form. In fact, the Anki card template I used for Japanese mining already embeds Jisho links for the tangos I learn and the examples sentence for the tangos. Jisho pages with tangos or example sentences will pop up after I click the tangos or example sentences.

[Asher]

It is mind boggling to think that people are a prisoner of their own language. The way your brain works is literally controlled by your native language, and that is a scary thought not because of the ‘control’, but because of the endless potential that is hiding beyond that small fraction of the world experience that people live within.

[Cassowary]

It sounds like an amazing tool, and I’m sure that as AI develops it could certainly streamline learning the language, but I feel I should mention the caveat that, currently, I don’t believe AI (at least not OpenAI) works very well in Japanese.

My partner, who is a Japanese university professor in Japan, tried to use OpenAI in order to help her with both translation work and streamlining writing and formatting business emails to cut down her workload, and she’s very disappointed with OpenAI’s ability to work with the Japanese language currently. She’s found it very useful when working in English, but continues to complain about its performance even working with simple Japanese.

Of course, this is just my 2-cents based on second-hand, anecdotal information, but I just thought I should put it out there!