There’s an interesting disconnect here, where the author of this piece apparently wants to learn French, and knows English, and they want to share how they went about learning French with an English audience.
But despite being a native English speaker, they couldn’t be bothered to actually write the article themselves. What’s the point of learning a language - any language - if you’re not even going to use your own words like an adult?
I think the author would have had a better reception without the triumphant “canceled French tutor” framing. The juxtaposition of “learning a language” with “cutting out the human element” is very off-putting.
Yeah. I am definitely not saying one must spend money on tutors. You can learn French without tutor, actually. French is one of those languages that have fairly large amount of free or cheap high quality resources available online.
But, the learning writeups that lead with the "learning something new" part kind of feel better the those who put unnecessary emphasis on "canceled tutor". So, you get something inspiring like a book they can read now (and could not before), about podcast they understood (and could not before) or a little about successful small talk in French. I have no idea about what the author learned in what time.
"I, a person who does not speak French or understand its grammar, am convinced that AI has made me better at speaking French and understanding its grammar than a tutor would."
This just isn't a reliable source, even putting aside trying to measure language proficiency through memorising grammar.
LLMs surely can help with language learning, but when they write posts like this, the zombie/body snatchers/borg effect is so strong as to be unreadable. Just write it yourself! You can do it! It will be better, please please stop generating bad boilerplate language with these fascinating algorithms, PLEASE!
I am doing something similar but mix more simpler and didn’t even get to the stage of building an llm tool. I am just using ChatGPT to distill common learning books and set aside 1 hr thrice a week. I made considerable progress, to me and then I used Italki to confirm.
I used voice mode on ChatGPT to learn the tones for mandarin, and general vocab and sentence structure while for Japanese it helped me expand proper sentence structure greatly.
It sounds silly, but it helped reenforce a base structure that is helpful and having it confirmed by a tutor was nice. Best is I can really do it whenever. What op posted does sound next stage, and I can imagine it’d be a viable platform.
I don’t suggest notebookllm to make an audiobook, I tried and it was the most dryest speech I ever heard. It did sound convincing enough if you were to do a podcast for it and that is what it does.. but it was completely horrid for learning but maybe that’s just me.
One of the cool things about LLM-assisted language learning that I have found is this:
Language is (for almost every adult) deeply personal. Even the best of teachers must be so nuanced for every error correction, repetition request, nudging, encouraging, etc. Why? Because the adult student is greatly affected by human feedback in this context.
This is one of the (many!) reasons why children learn languages (their first included) kinda fast. Their ego isn't involved.
Learning with a non-human, at least for me, is kinda cool as I don't feel bad telling it "look, I get it, don't ask me that again" and I don't take it personally when it says "that's not quite right."
I've got tons of experience as both a language student and a teacher, btw.
Can you please explain what you mean by distilling learning books and learning from it? Just passing in a chapter and asking the model to help you work through it?
This was so obviously LLM-authored that I stopped reading after just a couple of sentences. The author here has done themselves a disservice - mastering complex written grammar is meaningless if you cannot speak it or recognize spoken language. Interacting with a real French person for a few minutes should have been enough to cement this.
I complain about LLM-slop a lot (see my comment history), but this one doesn't feel like one!
The only bit I thought could be LLM was:
> The gap in my tutoring setup wasn’t the quality of the instruction. It was two structural things.
Even here, the second sentence is quite weak, none of that stupid punchy sentence that LLM likes to copy from (probably) TED Talks because it thinks that'll make the text fucking profound.
Watch movies and listen to people if you want grammar to stick. Languages are living things. Not something you practice in a bubble with Anki and Duolingo.
There’s an interesting disconnect here, where the author of this piece apparently wants to learn French, and knows English, and they want to share how they went about learning French with an English audience.
But despite being a native English speaker, they couldn’t be bothered to actually write the article themselves. What’s the point of learning a language - any language - if you’re not even going to use your own words like an adult?
I think the author would have had a better reception without the triumphant “canceled French tutor” framing. The juxtaposition of “learning a language” with “cutting out the human element” is very off-putting.
Yeah. I am definitely not saying one must spend money on tutors. You can learn French without tutor, actually. French is one of those languages that have fairly large amount of free or cheap high quality resources available online.
But, the learning writeups that lead with the "learning something new" part kind of feel better the those who put unnecessary emphasis on "canceled tutor". So, you get something inspiring like a book they can read now (and could not before), about podcast they understood (and could not before) or a little about successful small talk in French. I have no idea about what the author learned in what time.
"I, a person who does not speak French or understand its grammar, am convinced that AI has made me better at speaking French and understanding its grammar than a tutor would."
This just isn't a reliable source, even putting aside trying to measure language proficiency through memorising grammar.
LLMs surely can help with language learning, but when they write posts like this, the zombie/body snatchers/borg effect is so strong as to be unreadable. Just write it yourself! You can do it! It will be better, please please stop generating bad boilerplate language with these fascinating algorithms, PLEASE!
It's hateful to read.
I am doing something similar but mix more simpler and didn’t even get to the stage of building an llm tool. I am just using ChatGPT to distill common learning books and set aside 1 hr thrice a week. I made considerable progress, to me and then I used Italki to confirm.
I used voice mode on ChatGPT to learn the tones for mandarin, and general vocab and sentence structure while for Japanese it helped me expand proper sentence structure greatly.
It sounds silly, but it helped reenforce a base structure that is helpful and having it confirmed by a tutor was nice. Best is I can really do it whenever. What op posted does sound next stage, and I can imagine it’d be a viable platform.
I don’t suggest notebookllm to make an audiobook, I tried and it was the most dryest speech I ever heard. It did sound convincing enough if you were to do a podcast for it and that is what it does.. but it was completely horrid for learning but maybe that’s just me.
One of the cool things about LLM-assisted language learning that I have found is this:
Language is (for almost every adult) deeply personal. Even the best of teachers must be so nuanced for every error correction, repetition request, nudging, encouraging, etc. Why? Because the adult student is greatly affected by human feedback in this context.
This is one of the (many!) reasons why children learn languages (their first included) kinda fast. Their ego isn't involved.
Learning with a non-human, at least for me, is kinda cool as I don't feel bad telling it "look, I get it, don't ask me that again" and I don't take it personally when it says "that's not quite right."
I've got tons of experience as both a language student and a teacher, btw.
Can you please explain what you mean by distilling learning books and learning from it? Just passing in a chapter and asking the model to help you work through it?
We've all seen how LLMs write, imagine someone who talks like that.
C'est un véritable game-changer...
Non fromage, omlette.
This was so obviously LLM-authored that I stopped reading after just a couple of sentences. The author here has done themselves a disservice - mastering complex written grammar is meaningless if you cannot speak it or recognize spoken language. Interacting with a real French person for a few minutes should have been enough to cement this.
I complain about LLM-slop a lot (see my comment history), but this one doesn't feel like one!
The only bit I thought could be LLM was:
> The gap in my tutoring setup wasn’t the quality of the instruction. It was two structural things.
Even here, the second sentence is quite weak, none of that stupid punchy sentence that LLM likes to copy from (probably) TED Talks because it thinks that'll make the text fucking profound.
Funnily the author already made a very basic French error at the end. Wrote "bon chance" instead of "bonne chance"
And as we say, _c'est bien de la merde_.
Watch movies and listen to people if you want grammar to stick. Languages are living things. Not something you practice in a bubble with Anki and Duolingo.