Hello everyone,
and welcome to your German Word of the Day. And today’s word is:
Spielzeug
Which of course means toy, but we’re not actually going to talk about the word.
I mean… it’s Spiel+Zeug, “play-stuff”. What’s there to explain :).
No, instead of talking about Spielzeug, we’ll actually use Spielzeug.
Because I have created a little toy for German learners.
And it’s with ChatGPT…
Yeah, I’m actually not kidding.
I’ve built an example generator with ChatGPT.
Now, many of you have probably heard about ChatGPT, but some of you have not so here’s a little primer. Feel free to skip it, if you’re up to speed about it already.
What is ChatGPT
ChatGPT, or the GPT-series is a very large and powerful artificial neural net by a company called OpenAI (which is of course completely proprietary). The model is built and trained to create coherent and sensible human language and it’s very impressive how far the tech has come in the last two years.
Lex Fridman and other people who larp as AI experts are even low key suggesting that there are some sparks of consciousness in it.
I think that’s complete nonsense and I have actually spent quite a bit of time probing the model and I have plenty of examples that show that there is absolutely “no one at home” inside this thing. Like, seriously… I can post you some screenshots of how it fails very simple tasks.
It’s just a very very very complex machine.
Still, it’s very impressive what we can do with it.
I actually had my eye on OpenAI since 2020 because I wanted to use it for language teaching and learning. It’s just logical, and Duolingo actually does that now.
Quick Duolingo Rant
They have integrated GPT4, the most recent and most biggest version of GPT, in their new, flashy Duolingo version called
Duolingo Max Price
Or Duolingo Max for short.
You can have dialogues with the AI now and you can ask it to explain why an answer of yours wasn’t accepted, which is cool. But the offer is 20 bucks a month. Which I find heinously overpriced.
I mean… Duolingo says in its blogposts that they’ve done some super duper engineering together with OpenAI, but I don’t believe it. I think that a couple of engineers just did some fine tuning of the GPT base model while 90% of the team was busy A/B testing what the better prompt is after a lesson.
And now they’ll be getting swamped in reports about wrong AI sentences that they’ll never get on top of. Because… GPT is not all that reliable when it comes to language learning. It does make mistakes and spit out unidiomatic sentences and phrasings from time to time. I’d say 80% is good, 20% is not idiomatic. So for millions of users, that makes hundreds of thousands of possible reports per day. And no, I do not believe that they retrain the model as they claim.
But then again, Duolingo has spent years making its users accustomed to weird examples that make no sense and seem random, so GPT fits right in there. They really played the long game there. Well done Duolingo!
Anyway, I think ChatGPT is not quite good enough yet to be integrated in actual reference material (like my dictionary for example), but I also think it’s a useful tool and I have a number of plans I want to use it for in the context of learning German.
Anyway, the thing that’s already done, at least for now, is… drumroll… my epic example generator
My New Example Generator
The idea is simple – you enter a word (or a short phrase) and you get an example in German along with the English translation. And because that alone is boring, I have added multiple style options and I’ll probably add more.
So you can generate an example for the same word in various styles :).
Here’s a screenshot from when I was testing the system:
I mean… that’ some real wisdom.
And this isn’t heavily cherry picked either. The Buddha quotes in particular are usually really good.
I find it really quite fun to play around with, but there are a few things you need to be aware of when using it:
1 ) It sometimes ignores your word:
Instead, it uses a relative or synonym. For instance, instead of leihen it might use borgen. And for Überheblichkeit, it kept using the adjective überheblich. The reason for that is that words are processed as concepts and those are somewhat fluid. If two concepts (or vector embeddings, for the techies among you) overlap a lot, there’s a good chance the neural net takes in one and puts out the other.
I could try and force the model to use the exact word entered but then it would probably not conjugate anymore or stop using verbs in various tenses.
2) It sometimes says things that are not idiomatic:
In one example, it used groß in the sense of great, awesome, which is absolutely not idiomatic German. So if you get something that strikes you as odd, chances are, it probably is. After all, the model has no concept what’s idiomatic and what isn’t.
3) It doesn’t always get the grammar right:
Especially when you enter a phrase like bestes Bier, it might use the phrase in a sentence without adjusting the adjective endings. I think this is something I can iron out over time by fine tuning the prompts I am using, but
4) It gets stuck on one phrasing
Sometimes, it gets hung up on one meaning or use of a word if you generate multiple examples with it. For instance, I tried out doch and once used it as aber in the first position, it then kept doing this again and again. If that happens to you, pick a new word or refresh the page.
5) It struggles with prefix verbs and fails at words like “doch”
Yes, the quality of examples for prefix verbs is comparatively low. Or let’s say, the failure rate is higher than for nouns for instance. And don’t expect to get anything idiomatic for words like “doch” or “halt”.
I will soon add an option for “context”, so you could tell it to make an example for ausmachen in the context of “light”, for example. I think that’ll improve it somewhat.
***
So yeah, these are the drawbacks of the system.
If you want to systematically collect examples for example for all the meanings of a prefix verb for example, then this is NOT the right tool.
It’s more of a fun to play around with, and collect some vocabulary and it works best with nouns and adjectives.
But enough with the words… time to try it out!
Viel Spaß!!
***
***
I’ll add this to a dedicated page soon, so you can always find it directly from the menu and it’ll be a members only feature then, but for now, it’s free.
And I have several ideas for additional features and I’m also working on making it into an actual app. I just don’t have experience with that, so step by step.
Anyway, if you have other ideas for features, let me know!
And also let me know how you like the tool and which style is your favorite. And of course share your examples if you have a really hilarious one.
That’s it from me for now, have a great week and see you next time :)
Oh, and if any of you have used Duolingo Max… I’d really love some impressions and how the AI is doing there. And while we’re at it… Duolingo is not the only company that has “closely worked with open AI”.
Also, if you want to play around with AI learning, check out Langotalk.org
It’s got the same features as Duo Max, plus some more, plus it’s cheaper and looks better.
Ok…I bit. After an all too brief demo time, I went ahead and signed up for Langotalk. I’m about to enter my Sommerferien so… I’m in!
My first experience with AI – take that with all the salt you need. Amazing. And odd. For Langotalk specifically, I love some of the features that help you actually LEARN the language. Being able to save words and phrases that “Anna” uses with me is great since they are a part of the conversation we are having (not just random phrases). Instant translations of only what you need – awesome. Being able to just talk with no typing, the best.
The first odd part is that on a few occasions, when I ask Langotalk (the app, not Anna specifically) about the grammar I used it gets something wrong. For example (just happened again), I told Anna (first name basis already…sigh) “Ich esse jetzt zu Mittag”. Since I am a novice, and I am literally just talking, I always query and have Lango tell me how I did. Lango said “Good job. However, since Esse is the first word in your sentence, you should capitalize it”. I double checked it’s translation of my spoken word, and it clearly says “Ich esse jetzt zu Mittag.” So…no idea what is happening, but it has done this a few times.
Finally…Anna. She really wants to know where I grew up. I would have no problem telling 90% of the world this (its online I’m sure), but her…she feels…clingy. Why does she want to know what city I am living in? What does my government security clearance have to do with anything? (last one, teasing). Anna gives me the sentences using Dativpräposition that I asked her for, aber…I feel she doesn’t want to do that. Honestly. She switches gears and wants to talk about where I’m going on vacation and when (!). She is clingy. I need advice on how to create a little ‘space’ in our relationship. Its me, not her…
That’s the one thing that has kept me from actually doing a post on Langotalk so far – the grammar advice is faulty and the “learn about the word” section contains mistakes or nonsense 50% of the time and lacks crucial information 70% of the time.
The main reason is that the developer has only “optimized” the program for Spanish, which I take with a huge grain of salt because I do not believe that he has fine tuned a gtp3 version. I think he just refined some prompts for Spanish, and then translated them, but my suspicion is that also the Spanish grammar and “learn about word” sections contain mistakes.
It’s not Langotalks fault, it’s just not something GPT is good at and it should not be used for these kinds of instructions in my opinion, at least not unsupervised. Just too many mistakes that confuse learners.
I’ve done a fair share of testing and “research” to get a sense of what the gpt machine can and can’t do because I wanted to use it as well, but grammar feedback is not good enough imo to be used in a product. But that’s why I am not rich :D. Other people will use it, no doubt, and customers will find value in it.
As for Anna being clingy… I think the way the chats are “designed” is that the bot is constantly trying to get the initiative in the conversation, because then it’ll flow best. So they keep asking you stuff and depending on the “genre” they have a fixed selection of things to ask. This is not GPT, this is the preprompt that langotalk delivers along with the chat (my analysis, I don’t know this for a fact, but I know enough about chatGPT to make an informed guess).
I think the chat function is great, especially with audio. I also liked the “give me a story and ask me questions about it” section.
That alone is worth the money and worth more than a duo subscription.
The “rate my grammar” is to be taken with a boulder of salt, and the “learn about word” section is useless until a proper prompt is added by the dev.
He’s very active on discord and also added a few of my suggestions pretty much within a day, but he’s more or less alone, and I’m sure his to do list is a kilometer long.
Anyway, glad to hear you’re enjoying your experience there so far. GPT as a generator is a great tool for learning so long as you keep its limitations in mind. They will likely not go away anytime soon.
The oddest bug is when you enter nothing. Nichts. Press the ‘talk’ button is a quiet room and say nothing for a second, then stop the recording. It makes up very very random things and the AI then tries to answer. The last time it entered “Untertitel von Stephanie Geiges”. Anna started calling me Stephanie after that event. And I thought we were sharing babe…
These bots do really well so long as you stay “on script”. If you veer off, they might get confused a little.
Seems like you’re enjoying yourself though, that’s great :)
I asked for an example for verhalten and got:
Er zeigte ein völlig normal verhaltendes Verhalten im Büro und alle waren zufrieden.
He showed completely normal behaviour in the office and everyone was pleased.
So my question is I don’t understand verhaltendes verhaltend. Does this mean behaved behaviour?
Could you explan what verhaltend means in this respect.
thanks.
good system for examples by the way :)
It’s a good example for a bad example :).
The “verhaltendes” doesn’t make sense here and needs to go.
80% of the examples you’ll get are okay, 20% will have some form of mistake in them, infortunately.
Hallo: thanks so much to the generous donors that allow me to study. I very much appreciate your help. I look forward to conversing with my new in-laws in German.
This just does not work at all for me using Chrome. I do not even see the box where “let me think for a second” should appear, so I never see those words, and never get any output. I just see a small red circle with a red line through it.
Using different browsers I do see : “Let me think for a second”, but only once have actually got any output. I have reloaded multiple times.
John
Did you try clearing the cache? That’d be ctrl+f5. Sometimes, the browser is using old scripts and needs to load new ones from the server.
Thanks for this reply. Initially this did not seem to help, but today things are better.
In Chrome, my usual browser, it looks like it generates a sentence, but just can’t show it to me.
In Firefox it is working properly.
I assume I have an Extension that is getting in the way.
John
So the freezing was actually caused by a bug I fixed yesterday. Not sure about what you mean by “can’t show it”. Maybe it’ll fix itself in time :). These things are usually a bit glitchy at the start.
Thanks
What I mean by “can’t show” is this.
In Firefox, beneath the “Generate Example” button is an empty rectangle. when I click the button, that is where I see the “let me think a second..” words and then the output.
In Chrome that rectangle is not present, therefore I cannot see what should be in that rectangle. I imagine the system is trying to show it to me, but it can’t.
John
That’s a REALLY strange bug, I have to say. Is that on desktop or on phone?
This is on a Windows 10 computer.
I have now turned off all my Chrome Extensions, and the missing rectangle has appeared, and the output now appears as it should.
I don’t which extension was the problem. Maybe DuckduckGo Privacy.
Hard to tell! The only thing that I know it could be is Javascript, but lots of websites would have issues if you blocked it.
Anyway, good it’s working now.
And just for information… there are zero trackers in that code. The only thing I am doing is taking your word and sending it along with a few added lines to the model. No IP address or userinfo gets seen, so I wouldn’t know why a privacy plugin would act up.
Ich finde das sehr cool. Es ist interessant für mich. Ob es (KI) die Wahrheit ist oder nicht, weiß ich nicht, aber es ist ein guter Partner.
Ich habe Sie auch im Easy German Podcast gehört. Immer eine gute Zeit!
Could not come up with a poem using Esslöffel! (LOL!) Or maybe my computer is just lagging. Cool feature!
I think GPT was lagging, not your computer. That happens sometimes :).
Here’s the poem I got just now:
Der Esslöffel schwingt
In meiner Hand,
Wie ich es liebe
Es zu fühlen.
The spoon swings
In my hand,
How I love
To feel it.
Well Matt – I tried your AI search thing, with the word “Schwaden” (Normal). After 1.5 hours – nothing. Still wants to think for a second . . .
I just tried it and it worked quickly. I think GPT might get hung up sometimes, but it’s not really dependent on the word you put in. Just reload when this happens.
I tried with Erfolg.
I like normal,Nietzsche, Buddha ,arhaic and formal style. It’s a great tool ,useful and funny.
Duolingo was somehow appropriate for passing the boring frightfull Covid years but I would never dare to compare it with your level and complexity.
Of course for a Scorpio as you are ,you feel the need of crushing it but believe me ,this isnt’ t a fight worth for you who are far better !
Hahahaha… that was a nice comment :). Vielen lieben Dank, ich habe mich sehr gefreut!
Help me graduate! / Hilf mir zum Abschluss!
Dear community,
I’d like to invite you to take part in a survey concerning the Russia-Ukraine war for my bachelor’s thesis.
https://www.soscisurvey.de/BachelorThesis_ChenXi/
Takes only about 5-7 mins
Thank you!
“Lex Fridman and other people who larp as AI experts”
lol sehr lustig
2 typos:
“And now THEY’LL getting swamped in reports about wrong AI sentences”
they’ll –> they’re
“…but I also think it’s a useful tool and I have a number of plans WHAT I want to use it for in the context of learning German.”
what –> that
And my fortune, after I immediately reverted to a juvenile and used ‘Scheiße’ as the prompt:
“Ich habe mich überall mit meiner Scheiße beschmutzt und bin jetzt völlig erschöpft.
I have dirtied myself with my shit everywhere and now I’m completely exhausted.”
Ok a little disconcerting it knew exactly what I did today??
Wow, what a descriptive example :).
I of course also entered all the basic “basic funny words for 13 year olds”.
My favorite was this:
– Es ist besser ein Arschloch zu sein, als ein Arschloch zu haben.
Really not sure if Buddha said that :D.
Duolingo didn’t teach me this (yet?) for the meaning of “ein und zu,” but the Chat Buddha Quote says,
“Ein und zu können wir die Welt nicht verändern, aber wir können uns verändern, um die Welt zu verändern.
Alone and together, we cannot change the world, but we can change ourselves to change the world.”
Ich mag das. Ich weiß nicht, ob das Deutsch richtig ist, aber ich mag es.
Chatty clearly didn’t really know what to do with “ein und zu” and interpreted it as “zusammen”.
Note that it CAN’T give you an example that would show the different uses of “ein und zu”. It understands that it should make an example that contains this phrase, but it doesn’t understand the “meta” implications like a human would because I’m not prompting it to do that.
It’s quite amazing what all you’ve incorporated to the website. This is a one stop for any German learner!
I would surely be suggesting this website to my German teacher who’s going to pursue C2… I’m sure everyone , be it rookie or an advanced learner always has something new they can learn from YDG!
Again, great job Emanuel!
Vielen Dank :).
It’s not exactly one stop shop maybe, but I’m trying to make it a proper “homebase”.
Your lessons have helped me out a ton; thanks for the work you do. I like Duolingo for a fun practice tool, but it can’t get anywhere near the depth of a nice, deep grammar dive.
Danke :)!
I’ve been using Duolingo for more than 8 years (currently with a 7+ year streak) and have seen many changes. The loss of Discussions is sad as I found it helpful in understanding mistakes. I suspect many of the features that I used pre-Max have or might reappear in the paid version. Seems like a thinly veiled nudge to sign on to the paid version. I agree with Emanuel, the paid version is overpriced.
And here I was, feeling proud of my 3+ year streak at Duolingo. :-) It says I’ve been a member for seven years, but I don’t remember that. Congrats on seven years. I took German in high school and did well. It’s been a lifelong endeavor to learn it after I made a promise to myself after my much-loved Aunt from Germany passed away.
I offered to help alpha or beta test Max, but I never heard from them. I don’t think I’ll get that. However, I do have Chat GPT 4.0 subscription… I feel like playing with this. Would Emanuel release the code? :-)
Hmmmmmmmm…. not sure if I want to do that :). BUt if you tell me what you want to do, I can give you some pointers.
Just had a question about Zeug. Play stuff doesn’t make sense because in English a toy is singular and stuff is always plural. “Play+thing “would be a better literal translation no? Zeug, as i understand it can be both plural and singular in german depending on the particular use of the word?
So Zeug is always a singular noun, but it almost always refers to a collection of items, EXCEPT when in compounds like Werkzeug, Spielzeug or Feuerzeug.
So you’re right, plaything would be the more accurate translation, but I always feel people then take away that “Zeug” means “thing” by itself, and that’s not the case.
Thanks, Emanuel; that was helpful because I was thinking of Spieldinge or just Dinge, but the comment helped clarify… I’ve never seen the word Spieldinge; perhaps that would refer to a special friend?!
I’m a longtime paid subscriber to Duolingo, over 4 years. I’ve seen it go from bad to worse as Pearson took over. In the last few months the German to English translations of Duo are getting more bizarre than ever. Also, Duolingo is not using synomyns. It always interprets ‘viel’ as ‘a lot of’ and ‘zig’ as ‘a ton of’ (which I really dislike). German word usage is also suspect. I’ve compared their usage of German words to that in dict.cc, Oxford Duden, and Collins and it appears that their suggested usage is often uncommon. Worst, Duo discontinued discussion of the lessons. The ‘Discussion’ button is still there, but click it all you want and it does nothing. Talk about adding insult to injury! If discussions were available, these problems could be explored and native German speakers could comment. No way I’ll sign up for Duo Max.
Are you really that unhappy with the current version? The new Snake tree is working well for me. I hit a block with the previous version and started picking up Spanish. But, since the new version came out, I have been able to make better progress. I’m not using Max. I thought about it, but I’ll pass for now.
Three years ago, I was constantly commenting on the discussions. Some sentences were ridiculous. Now I think the German is better, and I question the English. I don’t know that I know enough German to say it has problems too. I know the English sentences do. Duo says, “Will we meet 3:00 pm?” I would say that should be, “Will we meet at 3 pm?” Maybe Duo is proper grammar-wise; I suspect NO! But it sounds terrible to my ear, which has always been the moderator’s excuse and justification in discussions, saying, “No, that’s wrong. I’m a native speaker, and that doesn’t sound right.”
I started taking the German tree, just for review purposes, and I was abhorred at the examples. It is a BILLION dollar, publicly traded company. We really should expect more. It’s not like some 10 people startup.
They proudly talk in their blogposts about how they found out through “research” what the better tagline is after someone completed a lesson, and all the while, lesson contain sh*t like this:
Or even worse:
Having idiomatic, grammatically correct examples in the in beginning of a tree for a language pair like English-German is the absolute minimum a paying customer should expect, in my opinion.
As long as they don’t clean this up, I do not respect them at all. (and I have more criticisms)
OK, OK, I’ll be the one to bite: apart from the evident and deliberate “silliness” of the first example(**), and my suspicion that the second one should end “…um ein Geschenk zu kaufen”, what else is objectionable here? Maybe it’d be fun if some “non-Emanuel’s” weighed in with their analyses!
(**) I have vague memories of claims that the use of such weird scenes (everyone has their favourites: “Haben Sie wirklich zweihundert Kartoffeln Ihrem Koffer?”, etc etc, ad nauseum) lead to better retention of vocab/structure, but colour me immer skeptisch…
The German sentence in the example is correct, but the English is not. A billion dollar company can have a whole team dedicated to correcting errors, but they don’t care.
Hmm, I’m not so sure about the “error” claim (even if the claim that they could afford a whole team dedicated to correcting errors is bang on the money): I understand the objection, but it’s a relatively common construct. “I’m going to the swimming pool and swimming a few laps”, or “I’m going home and spending some time with the kids”, or “we’re running down the sand dunes and jumping straight into the water”, or ” because of the boss’s unreasonable demand she’s going back to work and printing out a file”. Purists will definitely be able to file objections here (“I’m going to the swimming pool *to swim* a few laps” usw), but descriptivists be-a-having a field day, with both thumbs up ;-)
Hoping some other folk will weigh in here!
Oh wow, okay, then that was a big fat self own. I have never heard this type of phrasing I think and it sounds and feels viscerally wrong to me. Actually, it even feels like a mistake a German speaker would make in English.
I don’t like Duolingo’s sentence. It doesn’t sound good and I can’t picture myself ever saying that.
But pmccann’s examples with “going” sound just fine. I want to say that’s the most common verb you’ll see there.
It basically acts like a future tense. You could use “going to” in all of those examples. “I’m gonna go home and go to bed (probably pretty soon after I walk in the door)”.
It’s colloquial, super idiomatic, and kind of tricky to get right.
Hold it, you are talking about proper German and English and open with ‘big fat self own”? That dog doesn’t hunt…sorry.
If we are talking about the English in the second example, this phrasing and structure is really standard. If we are talking about 200 potatoes im Koffer, or a horse in the apartment or Nina the witch and her six kids living in the castle in the woods (Damn, I should have done that in German, I KNOW that sentence for sure!), then that is something different. It is a tool.
It is ALL a tool. AI/KI is a tool. YDG is a tool. Flashcards are are tool (don’t jack with me about the singular and plural there…). More to the point is that Emanuel should be thanking DL for this; lack of clarity drives people to YDG for answers and a greater understanding of the language.
Finally – I’ve said this before, but, the app and website of DL are crippled compared to the iPad version. If you want to beat up on DL, use the iPad version. So many more opportunities to listen ONLY or dictate the answers (auf Deutsch). I know Emanuel hates all Apple products too, so maybe that is salt on the wound.
“Big fat self own” is fine, you just need to recognise/recall that Emanuel likes to shift frequently among different registers, whether in English or German. In particular, if you think of a 1990s/2000s gamer talking about “owning” someone else (OWNED!/PWNED!) with a cutting remark or a killer move, you can see where this expression originates. I guess I’d hyphenate “self-own” here, but I’d also be more likely to talk about kicking a “big, fat own goal”.
I also have to (strongly) disagree about the iPad app being superior. I just tried it again for a couple of lessons to check out a recent version (it’s been a while), and the keyboard shortcuts are bad-to-non-existent, the brain-dead bullshit with “hearts” drove me batty in about ten minutes, and the ads are utterly unbearable. (Yep, you can pay to make those last complaints evaporate, but I’m sufficiently –and continuously– grouchy with Duolingo that it feels like a point of pride to remain with the free version.) So, yeah, Duolingo in a web browser with a decent keyboard and an ad-blocker is like heaven compared to the iPad version for me. As always, YMMV…
Yeah, I meant that my choosing specifically this screenshot and non-mistake basically backfired and invalidated part of my criticism :).
Key commands…You can just talk into it, both in English and in German for many of the answers. Only the ‘put the words in order thing’ makes it otherwise. Most of my time is spent on the listening exercises (which I described in detail elsewhere on this site). It is a tool to be used, not for it to use you. I cannot speak for ‘the game’ aspect of it.
Using DL without paying is like using YDG without registering/paying. Who wants to only read the first part of Emanuel’s thoughts on a topic before you have to stop? No one. Gotta find out how Maria and Thomas are getting along this week!
Hmm, I don’t think that’s a very good comparison: I get maximum benefit out of using DL, there’s basically nothing extra of value that gets added when you pay (and I’ll not magically become 4.23453429 times more likely to complete the course: I’ve already done that, and more than once as they have changed the content). The paid version just suppresses most –but not all– of the most irritating aspects of the experience. Pay now and we’ll stop annoying TF out of you every thirty seconds! Which is not the YDG model at all.
Anyway, different strokes for/of different folks, and all that…
“lack of clarity drives people to YDG for answers and a greater understanding of the language.”
That is true, I suppose :)
But I disagree about “use the iPad version”. I am using the version for the devices that I have and if they don’t care to provide a similar experience simply because the big majority of their paying customer base has iPhones/iPads, then I can absolutely judge the product on the merit that I see. After all, they decided it’s “good enough”.
Yes. I guess I was lucky that I started out on the iPad and didn’t know any different until I decided to (first) use the website, and the the iPhone. I won’t defend their decisions, but there is a difference and I have described that in great detail elsewhere on this site.
If they’re so clever with their absurd examples, then they should at least be running threads that reoccur.
Like… some person has 200 potatoes in a suitcase – there’s a lot of examples you can build around that premise that could pop up here and there.
I don’t think they have thought all that deeply about any of their examples.
(Ugh, not sure what I was on while posting that: “non-Emanuels” and “*in* Ihrem Koffer”.)
And as for the ChatGPT Duolingo max thing… the project Langotalk.org is run by like one person, maybe 3, and it offers MORE than Duolingo Max already, for a lower price.
Duolingo is just acting like they’re using special tech there when in reality, ANY programmer can code up these features in a couple of weeks.
take a look at Russell Stannard, teachertrainingvideos and see what he suggests for teaching. brilliant ideas.
Great suggestion! Looks good on first glance and I’ll definitely check out a few videos.
Well that’s a lot of fun! Naja, das macht viel Spaß!
Ich mag die verschiedene Stile, aber die Witze sind komisch. Es gibt nur der erste Teil des Witzes, e.g. “Warum entschied sich ein Haus, in ein anderes Haus zu ziehen?” aber die Pointe fehlt!
Trotzdem ist es sehr cool!
Hallo,
Here are some typos:
“You can can have dialogues” (extra can)
“words are processes as concepts” (words are processed as concepts)
“take in one and puts out the other” (takes in one and puts out the other)
“in various tense” (in various tenses)
“the quality of example for prefixes verbs is comparatively low” (the quality of examples for prefix verbs is comparatively low)
“member’s only” (members only)
“plenty of idea” (plenty of ideas)
“if you have other idea” (if you have other ideas)
My AI is detecting an s-problem today :)
Cool feature! I tried archaic style for the word “Scheiße” (I know, it’s the child in me…) and got:
“Scheiße, was ist mit der Welt geschehen?”
Well, not sure, but actually fits in well with this “Space Odyssey” trend!
Bis bald!
What do you mean with “s”-Problem :)?
You keep forgetting your “s”s :) :)
Just a quick one: I’m pretty sure that Duolingo Max isn’t available for German_from_English, but only for the Spanish and French courses (for now), and only for those accessing it via apps on iOS. Doubtless that’ll change before too long, but can’t say that I’m that excited about the prospect of using it, having looked at their blog posts regarding its functionality. I guess it’s always possible that the speech bots will be an order of magnitude or two better than the old ones? That’d be worth *something*, I guess: getting the old mouth and throat and tongue moving in pseudo-conversations!
I understand that ChatGPT is not good at translating or speaking in other languages yet. Although they are training it, they’re saying there isn’t the same volume of data for it. That was the polite way of saying that the subset of other language speakers is smaller than the set of English speakers. ;-) Nevertheless, it is a big concern. The hope is that AI proficiency in other languages will help bridge the world divide. I’m all for that too! China may be the first to have a proficient AI in Chinese; unfortunately, it may not be proficient in English. (sigh)
Oh quite the contrary. It’s REALLY good at translating and speaking other languages (the common ones, anyway).
Translating works, because things are internally stored as multidimensional vectors which it can map to words in pretty much any language that it has trained with.
I think Duolingo is fine tuning their instances of GPT, and they’ll have to do that differently for each language, but it’s not really a GPT issue, but a Duolingo issue.
Das ist einfach super geil! Aber für mich hat es funktioniert nur mit Safari, nicht mit Chrome. Es hat auch einen einfachen Grammatikfehler gemacht (“Wenn du die Schwindigkeit beherrschen willst, müssen du den Wind besiegen.”
Oh, das ist ziemlich schlecht :D.
Aber was ist denn Schwindigkeit? Das ist kein Wort.
As far as I can tell (ie, this is a guess!) it has somehow divined “Schwindigkeit” from the hyphenation of various compound nouns involving “Geschwindigkeit”: cool words like “Konversierungsgeschwindigkeit”, “Reifenhöchstgeschwindigkeit”, “Produktionsgeschwindigkeit” usw: I guess it’s fair to say that the poor old hyphenation mechanism for German texts occasionally has a really, really hard job to do! But yeah, the conjugation error above is not exactly confidence inspiring.
On a brighter note, inspired by this post, I dialled up ChatGPT (just the free version) and asked it to chat with me in German, and to inform me of any mistakes that I made along the way. 15 minutes later I had a decent, if somewhat stilted conversation under my belt. We decided to “siezen”, it asked me what my plans were for today, and gently corrected a few infelicities. The experience was actually a lot better than expected, and I’d recommend anyone to give it a whirl.
This conjugation mistake is actually the worst I have ever seen GPT make. Not sure what happened there, but I’d love to see the input word.
All in all, I think ChatGPT is a great tool to just practice some basic conversations and integrating Whisper (the speech to text AI of OPenAI) and a service like 11labs (text to speech, with ANY voice based on a 2 minute sample), it can make for an offer that wasn’t possible a year ago.
You just can’t rely too much on the model correcting your stuff and explaining the rules. Maybe think of it as a B2 language student. What you’ll get is mostly correct, but not always.
The error with the AI may have occurred because I noticed that Google Translate understood die Schwindigkeit to mean die Gechwindigkeit. IOW, the speed. And fixed it to “If you want to master speed, you must conquer the wind.”
Knowing that Google understands that word even though it is wrong, there were probably datasets used to train Google’s AI that contain that word. Thus the bane of AI. It only knows what it’s been taught, and just like a human, it makes mistakes.
Google Translate is a valuable tool. Not perfect, but a beneficial tool. Therefore, I double-flip the translation whenever I use Google Translate to write and translate any language, even when trying to learn German.
Suppose I am writing in a foreign language. In that case, the point is to double-translate foreign=>native=>foreign=>native. If the native doesn’t change, I have a higher probability of having written it correctly in a foreign language.
Now suppose I want to translate from native=>foreign. The process is slightly different. I will translate native=>foreign=>native if the concept or idea is off or wrong in my native language. I rewrite the concept using different terms in my native language and repeat the process native=>foreign=>native until I can double-translate, and the concept or idea remains intact. If the concepts return to me intact in my native language, I get that same probability.
I love to play games with players logging in from around the world. I use this method and teach it, and I have discovered that it works well according to foreign speakers and from my own experiences and vice-versa.
I did that with the Buddha Quote mentioned at the start of this thread, and Google fixed the error in German. :-)
It was a terrible Buddha quote, and I believe the AI must have been getting advice from The Flash instead of Buddha! :-)
Here is the test. Hier ist der Test.
Der Fehler mit der KI ist möglicherweise aufgetreten, weil mir aufgefallen ist, dass Google Translate Schwindigkeit als Geschwindigkeit versteht. IOW, die Geschwindigkeit. Und stellen Sie es auf “Wer Geschwindigkeit meistern will, muss den Wind erobern.”
In dem Wissen, dass Google dieses Wort versteht, obwohl es falsch ist, wurden wahrscheinlich Datensätze mit diesem Wort verwendet, um die KI von Google zu trainieren. Also der Fluch der KI. Es weiß nur, was es gelernt hat, und genau wie ein Mensch macht es Fehler.
Google Translate ist ein wertvolles Tool. Nicht perfekt, aber ein nützliches Tool. Deshalb drehe ich die Übersetzung jedes Mal um, wenn ich schreibe und mit Google Translate übersetze, selbst wenn ich versuche, Deutsch zu lernen.
Angenommen, ich schreibe in einer Fremdsprache. In diesem Fall lautet die Übersetzung fremd=>heimat=>fremd=>heim. Wenn sich die Muttersprache nicht ändert, habe ich eine höhere Chance, es in einer Fremdsprache richtig zu schreiben.
Angenommen, ich möchte von nativ => fremd übersetzen. Der Prozess ist etwas anders. Ich übersetze nativ=>fremd=>nativ, wenn das Konzept oder die Idee in meiner Muttersprache falsch oder falsch ist. Ich schreibe das Konzept mit anderen Begriffen in meiner Muttersprache um und wiederhole den Prozess nativ=>fremd=>nativ, bis ich doppelt übersetzen kann und das Konzept oder die Idee intakt bleibt. Wenn die Konzepte in meiner Muttersprache intakt zu mir zurückkommen, habe ich die gleiche Chance, es richtig zu machen.
Ich liebe es, Spiele mit Spielern aus der ganzen Welt zu spielen, die sich anmelden. Ich verwende und unterrichte diese Methode und habe festgestellt, dass sie laut ausländischen Sprechern und meiner eigenen Erfahrung gut funktioniert und umgekehrt.
Ich habe das mit dem am Anfang dieses Threads erwähnten Buddha-Zitat gemacht und Google hat den Fehler auf Deutsch behoben. :-)
Es war ein schreckliches Buddha-Zitat und ich glaube, die KI hat sich von The Flash beraten lassen und nicht von Buddha! :-)
I am learning German but I am not yet so good as to think I could have translated what I wrote into German. How well did Google Translate do with German? Probably not perfect like a German speaker, but can it be understood the way I intended?
Seriously does it? Because even after three years of German, I see a lot of words I don’t know… yet! :-)