Ugh. You can just feel the management layers in this.
Management wants to do something with AI. It's so hot right now. Your product guy got an idea for a cool new feature. Tests show that users actually like it. It's helpful. Great.
But now you need to explain how you made Duolingo money. First idea, your product has a paid tier, add it as a feature to the paid tier too boost paid user retention and new signups. Great, metrics! But you're pretty sure that, while customers do like this when you ask, you're probably not moving the needle. What to do? Well, you're a mobile game company, focus on what always works: go after the whales.
Now your product is an exclusive EVEN HIGHER tier. Get the art guy to make, like, an Amex Black Card but for the Owl. Make him do a superhero landing or something. Run a bunch of ads at the paid users to convince them that they'd totally be helped by this. They'll be focused on the users who already pay Duolingo money, so those are perfect advertising targets. And you'll be able to show real dollars that your feature is pulling in. Hooray! That VP spot is as good as yours, I bet!
I mean, sure, that's more or less how a company operates. You make a new product, you charge for it, you advertise it. What's the big issue? This all seems pretty standard.
> The issue is: from beginning on and for a long time Duolingo advertised with "Learn a language for free. Forever." :-(
The original business model (from the guy who made Captcha) was that the learners would translate text for others, for a fee. So the payment was labor.
Unsurprisingly that didn’t work out.
The alternative to charging was going out of business, so hard to vilify them too much.
They tried. They failed. They tried something new.
I wish they would have tried harder. After I finished Spanish in tried to translate some articles but the UI was confusing and there was no motivation like the learning bit had
Not only was there more diverse content when learning (though for that I can at least recommend Clozemaster), but now that some part of my mobile experience keeps defaulting to showing me wikipedia in German[0], I keep realising how out of sync the different languages of wikipedia are.
[0] a bit of a tangent from the article, but it's been irritating me for ages now.
I'm in Germany, but English is the top entry in Settings.app Preferred Languages, and I'm using iOS 16: highlight a word, e.g. "German"; in the context menu select «Look Up»; this text is all correctly in English; tap either the title or in the links list the wikipedia link; what do I get?
>
The alternative to charging was going out of business, so hard to vilify them too much.
> They tried. They failed. They tried something new.
You only advertise with such a slogan if you are really sure that you will never have to pivot to a business model where you want to have the user pay (even if paying is "optional" (freemium)) since advertising slogans have a tendency to stick. ;-)
Yup. Having and paying employees is a good that deflects from all criticism of price.
When I doubled prices in my coffee shop, at first people were unhappy, but as soon as I told them I threw out the coffee grinding machine and replaced it with 3 new jobs of "hand-grinders", people were overjoyed and immediately stopped criticizing the price increase. I was just making money to pay employees afterall.
When I had to raise prices again because all the hand-coffee-grinder employees suffered severe medical injuries, everyone understood that too.
.....
My point, obviously, is that hiring a bunch of employees to design ways to trick people into clicking more ads (sorry, ahem, improving user engagement) doesn't mean that the company making money is suddenly an obvious and unquestionable good that can be used as a shield from other criticism
Honestly, being able to build a company that creates livelihoods while simultaneously teaching people foreign languages goes against many of the historical and modern definitions of "late stage capitalism".
But I assume you're just using the super modern catch-all fad term of late stage capitalism that means nothing.
Also, from TFA: "Duolingo Max is available to everyone"
C'mon, let's make lingots (the Duo reward system) into a cryptocurrency. Reimplement the service such that people doing language lessons are the proof of work, and miners (learners) get lingots as their reward.
Nevettem. Sajnos bebizonyosodott, hogy errefelé nem nézzük jó szemmel az alacsony erőfeszítést igénylő szarkazmust. Hajlandó vagyok figyelmen kívül hagyni a törvényszegést a havi Duolingo Max előfizetés szokásos megvesztegetéséért.
In fairness to duolingo, nearly everything duolingo does is like this. They are extremely metric and process driven (for better and worse of actual users).
I can tell because when I tried to figure out the cost of the existing Super Duolingo, I was unable to do so. The https://www.duolingo.com/super page gives me lots of information about tiers and benefits but never mentions a cost. I'm sure telling you hurt metrics. Fuck those guys.
Thank you. I too wasted a good five minutes trying to find out what any tier would cost, and failed. I wonder if the USD prices are different from the CAD ones?
I'd be very interested to know exactly what metrics allow them to ignore the fact that their most-recent app redesign has been, from what I can tell, unanimously disliked and shown to frustrate their userbase.
This is exactly the example I was thinking of. There is a Reddit AMA where a user asks why they keep it even though it's universally disliked and the CEO essentially says "I hear you but DAU is higher on the redesign".
I think it's an example where metrics are taken too far. Yes, better DAU is good, but if you are alienating 100% of your power users and getting more engagement from your short-term, new signup users that's going to look great for a few quarters but kill your app over the long term as the community experts start telling everyone they used to recommend Duolingo to to go use something else.
I think they can't make it free (or with their old subscription model) because OpenAI APIs are pricey. It would quickly eat up whatever money they can generate. They also hope their product is getting THAT much better you want to pay.
Making features without any monetization strategy and abandoning afterwards because you can't even justify the salaries of the devs working on it is way better? It's like you're mad at modern product development, ignoring the reasons why.
Making a monetization strategy without understanding the feature is just as useless. If it's new you have to build it in order to understand it. You need to invest some time into the features first. If it doesn't stick then discard it early or iterate on it and use the information gained on building better features.
I use Duolingo, but only free. They've given me free access to their super tier twice now for a few days or a week each time.
I assume at some point they'll do the same to try to give me a taste of this new product and maybe it'll be good?
Either way I'll wait and see for a free option. I understand your skepticism. But I hope it ends up being a fun (and importantly: effective!) new way to learn.
Great, you've just boiled down how online services are making money from their applications. Company x online uses x popular tactics for making money off of the services they provide, shock.
It's a bit of a boring story really.
Yes AI is 'so hot right now' so why not leverage it? It's hot for a reason and maybe just maybe, it is actually beneficial to the product you provide, so why wouldn't you implement it. I can see another angle if they didn't where people would start to say Duolingo is starting to become dated, due to lack of features such as AI/ML.
This is the second time in the last couple of days that I've encountered a page hiding the pricing as good as they can.
I wanted to see what this Max-plan costs, so I go to duolingo.com. Nowhere is a "Pricing" link to be found. So I go to "Super Duolingo", because that costs something and therefore should have a price.
"Test 2 weeks for free", "Free | Super | Super-Family" with "Test 2 weeks for free", but nowhere what I'll have to pay after those two weeks.
Click on "Test 2 weeks for free": Submit your personal information for creating an account, no pricing information.
What is this? How much does the Super even cost?
Management to Web-Designers: "No, don't put the pricing, we're irrationally expensive", or what?
I find that so incredibly off-putting. Whenever I see a company intentionally hiding pricing information I move them all the way to the back of the line for serious consideration. If I don’t manage to entirely put the product out of mind I’ll usually look for competitors or ways to get the service without paying for it.
If they’re going to be user-hostile, become a hostile user!
I didn't see any pricing information while logged out either, but as soon as I clicked the free trial button and then logged in, the next page that appeared was pricing. It's $12.99/month for an individual plan billed monthly, $6.99/month for an individual plan billed annually, and $9.99/month for a family plan (2-6 members) billed annually. I’m in California in case that matters.
I have severely mixed feelings about Duolingo. I can't imagine paying more than I already do for the "Super" option. If you're trying to learn Spanish or French, it seems like a pretty full-featured app.
Unfortunately, I'm trying to learn Greek and every aspect of it is sub-par compared to other languages. It doesn't even read you questions correctly. (It uses flat intonation as if it's saying a statement.) There are no vignettes like in Spanish and French where you watch a conversation and then summarize what was said. They just throw in new concepts with no instruction whatsoever. So you see a word that you know is a verb, but the ending is one you've never seen before. There's no chart or examples, or even explanation that "now we're going to teach you past tense," or whatever. Just suddenly there's a bunch of new stuff you don't understand and you have to go to the Internet to look up what the hell is happening. God forbid you ask a question on the forum. The Greek mods are stereotypically rude and controlling. Nothing is ever the fault of the app, it's always the user's fault, even when native English speakers are telling them, "The English translation of this sentence doesn't make sense and is not valid English."
So yeah, I'm definitely not going to pay for an AI that is likely giving wrong advice anyway.
Not relevant to the discussion but possibly for you: I discovered this 100% free no-signup audio-only course for Greek through a previous HN post. Amazing approach (like Michel Thomas but more charming and modern and free!):
Of all the language apps I used Duolingo has by far been the least useful. I've spent hundreds of hours but took barely anything away (for FR, ES) If you must pay go to a proper app with structured learning, like babble or something language specific.
I’m Greek and I’ve never heard of LanguageTransfer before (it’s even created by another Greek, Michalis).
I used for a few months the free tier of Duolingo in the past, but it wasn’t to my liking and although it helped a bit, I can’t say it taught me Spanish (perhaps the paid version is better).
For sure I’m going to try the free Spanish course from LanguageTransfer ;-)
Thanks! I read about it previously on HN, but haven’t had a chance to check it out. I’m going to do that now. For what it’s worth, I have learned stuff with Duolingo, but it could be so much more (for me) if they’d get feature parity with some of the other languages. I get that Greek probably doesn’t have a lot of people trying to learn it, but it’s sad.
Before the UX update at the end of last year, I was trying Arabic (something like 2 years of daily lessons and I still couldn't manage the alphabet), Greek (felt worse than before the preceding UX change but I couldn't say why exactly), Esperanto (the audio sounded like it was done by a volunteer at their own home?), Dutch, German, and Spanish.
I'm still doing German, because I live here now, but I gave up on all the rest the day that update happened.
Oh my goodness the voices and animations are annoying. Whenever the young boy appears I have to mute my speakers.
I'm sure they're hyper-focused on "engagement metrics" to the detriment of all else, just like how old-school OK Cupid turned into just another Tinder clone.
This is funny because in Spanish, the young boy (Junior) and his father (Eddy) are both absolutely fucking hilarious. But I 100% get what you're saying because I did some of the French lessons, almost identical content and Junior came across as snooty. It really does go to show that choosing the right voice actor makes a world of difference. The Spanish casting is absolutely phenomenal but the other languages don't seem to quite nail it. Part of me wonders however if this is actually a cultural difference. Perhaps I personally find the Spanish ones phenomenal because I am personally more aligned with Spanish speaking values, culture, humour and communication methods than other more reserved cultures such as those of German and French.
Well, if they are so focused on engagement metrics then surely they would notice an active user drop since the big UI update? Both my partner and I were using the app daily and completely dropped it since then. Going from the reviews we are not the only ones. It is just such a clusterfuck and I'm still feeling bitter about it.
Personally I find the German voice actors super annoying. The Indonesian ones seem fine. Funny how this is experienced differently for different people :)
> Well, if they are so focused on engagement metrics then surely they would notice an active user drop since the big UI update?
My guess is the "school homework assistant" market is very different from the "self-directed adult learning" market, and growth in the former outweighs loss from the latter.
Whatever the reason (I can invent many hypotheses, but they're just stories I can't test), Duo has gone from the app I enjoy making time for to the app I only bother with to keep my language lessons a little more diverse.
I occasionally do Duolingo for a little Japanese practice but I don't think it does a good job there either. The Japanese learning experience seemed far worse than the Spanish experience. Maybe it's because I knew more Japanese and passed in to the higher levels, or because I'm more aware of the errors, but I think it's more due to the fact that Japanese is further from English and Duolingo wasn't designed around it. It is missing tools around Kanji and regularly "speaks" using the wrong reading (since characters can be pronounced differently in different contexts). There are no tools around learning characters or grammar, no short stories or other supplemental exercises, etc. The language pack feels like a bit of an afterthought rather than a core offering.
The translations already aren't great, and I can't imagine polluting it with some unvetted AI translations would help at all.
You shoud google for duolingo grammar notes. You'll find some for every course I imagine. The same people who originally created those courses made unofficial wikia or other pages and populate unofficial discords.
I'm using it for serious study in multiple languages. It's still plenty useful.
They are just not hosting the community interactions anymore. Probably because it was too difficult to keep it clean from abuse and they decided to be about educational technology rather than about hosting and curating social content.
I don’t understand. Why aren’t the notes in the app? I can already search for language information on the web. But I’d prefer it was just in the language learning app I’m using already.
Duolingo is driven by analytics. My guess is that they saw that almost nobody is using the notes compared to the app and it's too much work to maintain them. Also they have to keep them working for three platforms, which isn't all that trivial.
As a Duolingo user who has refused to upgrade, if they roll the two into one it might entice me, but not for another higher tier above what they've got. But I guess there's a subset of super-users who may be keen.
To be fair, Super Duolingo is pretty good. Dropping the Ads alone is totally worth the subscription if you're a daily user (it's the only app I use daily aside from Messages, Email etc).
I'm very curious about Max though ... the explain your answer is something I'm really interested in to help me understand the sentences I'm constructing (which is where I make most of my blunders!).
I've yet to try Super Duolingo. But I've also been tapering off my usage as they keep forcing to use the word bank instead of keyboard at an increasing rate. I've no desire to use the word bank, they seem to have no desire to let me use the keyboard. Unless that's buried in one of the premium options I see the writing on the wall.
Interesting, I am a Super Duolingo subscriber but I don't think that's one of the features. Anywhere I'm inputting a sentence I think I get the keyboard option, maybe I'm misremembering, but I do a lot of typing on my Japanese keyboard when I study.
They were clearly doing some A/B testing a while back based on sporadic complaints on r/duolingo. At some point I noticed it had solidified. The current setup is if the drill is target language to host language, you'll always get word bank. If it's host language to target language it'll push you towards word bank but you have a subtle thing to click on that brings back the keyboard. Sometimes it's called "make harder" or sometimes "use keyboard" (vs "make easier" and "use word bank")
Try supplementing Duolingo with Anki / Ankidroid. Every word that I learn in Duolingo, I put into Anki. I do all my reviews in Ankidroid on a small E-Ink device but it's actually designed for phones.
You can configure Ankidroid to require you to spell the words properly.
If you enrol in the classroom version of Duo (which I originally did to get back the old style tree) it is firstly, free. Secondly - you can be a single participant. Thirdly - no ads. I even forgot there were ads because I haven't seen one since I enrolled - in eluding the annoying self promotion ads for Super Duolingo. I have a 1631 streak at the moment, so it is saving me quite a bit of time not seeing ads. It is not my loop hole, so I guess it is just something I found by accident.
Maybe I'm wrong, but I would not be surprised if forum moderation was a volunteer activity. Even core functionality like developing the lessons, for less popular languages, was done by the Duo community and given very little support from paid staff.
Yes, but moderating and screening for abusive content can't reliably left to unpaid volunteers if it needs to be good enough not to scare off parents and educators. I remember that this was the death knell to a Lego online game: People were making genitals quicker and more inventively than anybody could keep up.
In education technology, many seem to hope to get that from AI: the helpfulness of a forum of volunteers without the abuse. The rise of the "unsocial network".
They still have the forum style "help me out", but there are some where the responses are really a huge miss where I could see this AI helping out more by giving better answers.
I also use it daily but I'm a bit worried I won't concentrate as much if I have unlimited hearts, have you noticed any decline in recall? But yes, the ads are super annoying... especially the same one to upgrade Duolingo puts in every time!
> I'm a bit worried I won't concentrate as much if I have unlimited hearts, have you noticed any decline in recall?
I did a trial and I noticed I spend more time with the App when I had unlimited hearts, so I got through more lessons a day. Less attention to a single question, but a lot more time spent on the app.
And that when it comes to actual recall, the mistakes don't count when it comes to learning. I don't really learn by recalling information, I learn by having stuff shoved into my brain with Spaced repetition.
The unlimited hearts thing fed the repetition cycle a lot more.
Just starting my second paid year - I didn't put up with the adverts for long.
Works far better for me without the hearts. Less time staring at sentences trying to make them perfect - more quick failing, learning why, iterating and drilling it into my head.
Also much more likely to spend longer in the app, maybe a couple of times a week.
That's a new term for me, but after researching it, it makes sense. What are your preferred sources of comprehensible input. After a few months on Duolingo I still find TV shows to mostly not be comprehensible, so not productive.
I do occasionally get through some french youtube with subs and I'm getting better at roughly following - I'm not just relying on Duo.
(I also have a grammar book to hand as well)
I have found that it works better for me not have unlimited hearts. I make my mistakes, and in I am not ready on a level then I have to repeat it. And either way I then do practice rounds to fill my hearts back up to full (I try to always leave the app with full hearts).
In the times I have had SuperDuolingo for free Have been annoyed that I can just keep plowing through mistakes (really: lack of knowledge), and don't have to really learn things before moving on with them... I just have to guess right once for each question.
But I do agree that DuoLingo is all about learning by making mistakes. I wish they had a bit more focused teaching for grammar, at least it would help me with the Hungarian. For a few items I have had to go outside the app for lessons on what I am just not getting by repetition.
I haven't, but I got sick of the ads pretty quickly and have been signed up for super for most of my ~10 month streak.
I try to spend 10-20 mins a day on it (learning Spanish) and actually do a lot of the "review" lessons which are good at drilling words. I am often surprised by my recall at phrases that I haven't seen in a while -- pretty sure they use the spaced repetition type algorithms.
My challenges now aren't so much on vocab/recall, but on verb conjugation and sentence construction / structure. I kind of know what to say but have trouble putting it together.
What I need to do is practice more actually speaking Spanish and interacting with Spanish speakers -- I had planned to spend a few months living somewhere in Hispanic America, but the 3 day-a-week RTO at my job has trashed that. I'll probably just have to settle for Skype-based chats/lessons.
I've had success blocking the ads using a pi-hole. You still see ads for Duolingo Super and other in-house offerings, but the obnoxious video ads go away.
I've been partnering Duolingo with ChatGPT for a while now. I'll ask ChatGPT "Why was this wrong? What was my mistake?" and it explains exactly what I needed to know every time.
I think this combo is going to work fantastically well.
I recently asked GPT-3.5 ChatGPT to explain the grammar of a Japanese sentence using an obscure informal verb form, and it gave an answer that was obviously wrong. I just tried again with GPT-4, and it gave an answer that was more plausible, but which is almost certainly also wrong, because it contradicts the much-more-convincing Stack Exchange answer I had previously found via Google. (I only asked ChatGPT the question as a test.)
So, it can be helpful for basic questions, but there are limits. The harder the question is, the more likely it is to get the answer wrong, and the harder it is for the user to identify when it’s done so.
This is interesting because my impression is that LLMs perform poorly on non-english languages[1]. This is likely 'just' a corpus size problem - but it's not like that is going to be fixed any time soon. That said, at least in language instruction there's a lot of meta material that will be helpful. I wonder how Duolingo is detecting when GPT-4 is hallucinating.
LLM outputs in Standard Chinese are fluent and coherent but the grammar style and choice of words are clearly non-native. I think the source training material just isn't as comprehensive as English, and contemporary Chinese Internet is packed full of latin abbreviations and obscure homophones/wordplay due to its evolution under severe censorship.
Awesome observation re: censorship. Last I looked into training sets, Chinese and English were the frontrunners by far in available corpus size, and benefit from a certain amount of government-enforced-homogeneity (a standard mode is taught in compulsory education).
Arabic is safe from automation due to its limited exposure to the open web and extreme variability of dialects, and Icelandic will remain obscure due to sample size and the fact that they mostly speak English online.
Arabic is also often written in the so-called Arabic chat alphabet / Arabizi online, so a lot of Arabic text available on the internet looks less like حروف عربية and more like “7ruf 3rabiye”.
Formal text wise Chinese is still a very comprehensive language. However in terms of user-generated content it has degraded severely especially in the past ~5 years. Internet slangs in English can be confusing too, but users have a choice in whether to use them or not. On Chinese social media using abbreviations/homophones has become practically baseline for certain common terms because these platforms do so much preemptive self-censorship due to the government's near-invisible red line that fluctuates constantly.
They don't perform poorly (depends on how much corpus training the language received). https://arxiv.org/abs/2108.13349. The important thing is you don't need an equivalent number of tokens for an equivalent performance. So even though GPT-3 only had 1.8 % french by word count, it can speak fluent french. There's positive transfer between languages.
Also GLM-130B trained on equal English/Chinese (200b tokens each. 400b total) speaks both just fine.
The Twitter thread you linked is just somebody pointing out poor performance in Tigrinya. That doesn't mean it performs poorly on all non-English languages. It's performing poorly on the languages with little representation in the corpus, but those are also the languages that are least likely to be learned on Duolingo. Duolingo users mainly learn languages like English, Spanish, French, German, etc. and the models do just fine with these languages.
For what it's worth, ChatGPT is able to hold a meaningful conversation in Czech (pretty difficult, latin script, ~11M speakers). I can almost tell it's a bot, but it could also be a PR person from a big corporation who can't even afford to talk like a normal person anymore.
The GPT-4 post/paper covers this: GPT-4 does other languages excellently, and performs on par in many other languages with GPT-3.5 in English. As people found that useful...
Does it actually say how much Duolingo Max would cost? I do not see anything in that article or on their website. I don't have the app, maybe it's already live and pricing is available there.
I deleted Duolingo years ago when it felt like I knew everything it had to teach me, but was still far from fluent in the language. I assume it's gotten better, and would consider trying this new feature.
I've been successfully using ChatGPT for learning German in the last month or so. It does a great job enriching my plain B1 level sentences with some fancy verbs or adjectives and pointing out my mistakes.
How does that work? Do you just tell it that you are learning and to correct your mistakes, then converse with it in German? Does it infer your German level from your own chats with it?
Yes, you can tell it that you are learning the language at the beginning of the conversation in plain English and ask it to correct any mistakes. You can easily start a bilingual conversation with ChatGPT, some random example here: https://www.reddit.com/r/russian/comments/112xgp2/practicing...
To show you what I meant when I mentioned "enriching sentences", I've just generated an example of a B1-level English sentence improved by ChatGPT.
User: Please improve this sentence: "Many people want to learn languages because they like to travel"
ChatGPT: "Learning languages is a popular pursuit because it enables people to travel and explore different cultures with greater ease."
It works similarly with other languages.
As for my German level, it wasn't evaluated by ChatGPT but by an external assessment exam and I only mentioned it to add some more context to the story. Frankly, I probably wouldn't trust ChatGPT's assessment in this regard.
> User: Please improve this sentence: "Many people want to learn languages because they like to travel"
> ChatGPT: "Learning languages is a popular pursuit because it enables people to travel and explore different cultures with greater ease."
How is this an "improvement"?
The user gave a concise sentence without errors. ChatGPT made it more verbose and added additional meaning that wasn't present in the original sentence (thus going beyond mere "improvement"). Even if it stopped before the "and", the ChatGPT version is worse: it uses more words for no benefit.
You are generally correct but the context is crucial here - in this particular case it was my intention to get a verbose fancy sentence out of a much simpler one. ChatGPT inferred my wish from the previous prompts in that chat.
Such use of ChatGPT has been useful to me as I'm trying to progress from intermediate to advanced level in my knowledge of German. However, there are numerous methods for learning foreign languages and it is understandable that you may prefer your own ways and disagree with mine.
The request is weird and doesn't have much to do with language learning, though. It messed up with rhymes, but I don't think Duolingo has anything similar.
Like, who on Earth, if learning a language (and not just needing a phrasebook) would ever want to remember foreign language words by rhyming them with a different language?
While truly etymologically related words make sense - such as "escribir" and "scribe", connecting unrelated words is IMHO a bad idea that would do more harm than good.
I currently try learn some Spanish and I had some trouble with Spanish "mirar" accidentally being close enough to Japanese "見る" (which absolutely aren't related, of course, just happen sound similar), so now and then my brain had farted ("hey, I'm speaking in a third language, other than my native and English!") and I could blurt out "miro!" instead of "mira!" or some similarly weird shit unintentionally mixing up two different grammar systems as my brain struggles to explain itself through the grammar I know about. Maybe it's just me, but I think unrelated rhymes and false cognates are a big no-no when learning a new language.
When I've asked ChatGPT (GPT-3.5) to explain me Spanish grammar, such as "explain me 'a' in the beginning of 'a tu hermano le gusta...') it did a decent job explaining me how it works. Factually it matched what I found later, cross-checking.
It does seem like the model picked up examples of pronunciations about which people are admonished, and presented them as tips of what to do rather than what not to do. This basic issue with negation also pops up a lot IME with Google's search infoboxes, so hopefully people are at least slightly inoculated.
Yeah, Spaniards pronouncing a lot of 's' and 'c' sounds as English 'th' or 'z' is called the ceceo, and it's commonly incorrectly attributed to Spaniards copying a prior king's lisp.
If you drive from Virginia to Boston, you'll probably hear 5 different pronunciations of "Grassy ass". I don't think any of them sound like how I learned to say "thank you" in high school Spanish class.
Bot is a problematic example, as it has many different pronunciations depending on meaning and region. But non on the common pronunciations have an 'a' (IPA), like Gracias.
True, I was simplifying a bit. Technically "gracias" is /a/ as you point out, while bot is /ɑ/. But /a/ doesn't really exist in General American English, so bot (or bra, as sibling suggests) is the closest on the IPA vowel chart.
What's wrong with it? I'm a native speaker of neither, but they do sound similar (of course, gracias being pronounced South American style, not Spanish style). Except the fact that Spanish and English "r" differ, but I think it's reasonable for English speakers to know that, once they learn to pronounce Spanish "r".
What is wrong with the example? Although I am not a Spanish speaker, naively it looks fine. The pronunciations written in English are very impressive and do indeed sound somewhat like the corresponding Spanish words (albeit of course incorrect! But most people don't know IPA).
No, it's absolutely terrible on every single one. For some reason it's telling you how to pronouce them with an English accent, or rather how an English-speaking person would pronounce it if they didn't speak Spanish at all.
> Adiós: "Adiós, my toes" - rhymes with "toes"
In this example "toss" would be much closer than "toes".
But yeah, this is awful. I don't know how this happened.
Good question, it's not that easy to find existing words that match that well. If you asked me to do this I'd probably try making up fake words that in my mind, at least would sound closer (and that is also non-trivial given the weirdness of English pronunciation).
Also, I'm no linguist, my only credentials here are being bilingual, albeit with an accent (lived in the US from 0 to 8 years old and the past ten years, and a south-american country from 8 up to my 20s). So I lack the terminology to describe what things sound like.
That said - Cola, I think is fine. The "o" in toes is pretty different from the one in adios. I'd say that maybe it rhymes with "adipose" or "lactose", except the accent goes on the last syllable. Gracias, isn't too bad, but maybe I'd say "fascias" (admittedly a relatively obscure word). Manana is obviously wrong - here I'd just make up something like "man-nyah-nah" (accent on the nyah part). For uno, I guess I'd just say it sounds like "boo-no" without the b. Tres is close enough (the T and the D are different, but it still rhymes to my ear).
That is the way language learning books were written before the days they could provide audio recordings. Now with modern technology it is much better to learn the sound system of your target language then try to find comparisons with your native language.
I feel like the words it gave only work for _spanish with an american english accent_. But then again it maybe depends on how you would pronounce the words for comparison like banana.
With GPT-4, I'm thinking if Duolingo has fundamentally been reduced to a game.
The key network effect left is pretty much a leaderboard that fuels their daily retention. So any existing game company could potentially create another gamified learning app as part of their game pools.
There are lots of educational games or apps around language learning. Nobody is as successful as Duolingo. Or as useful as Duolingo, at least in whatever Duolingo chooses to do.
I don't even think they think being compared to a game is a bad thing. What makes Duolingo the best in its field is the content and the varied exercises with a high quality.
Yes, the gamification is the major positive feature that gets people to stick to it.
But the exercises aren't varied or high quality. Try drilling a particular skill and you'll soon memorize all their sentences, to a degree that it trivializes the practice. Not to mention they're absolutely riddled with errors that strongly imply they don't even run the translations by a native English speaker. Voice exercises with numbers are simply broken in Russian. And all the comments are locked, with many of the complaints ignored.
I never run into those issues. I use about 6 courses, and the English sentences have always been grammatically correct.
There's nothing wrong with memorizing example sentences. Your brain is not a digital memory bank, memorizing is impossible without forming some deeper understanding on the fly.
Would love to use duolingo, but they have no interest in supporting iranian languages. I’ve never really seen a clear answer for why. They have about 200 million native speakers worldwide. Persian, Pashto, Kurdish are the largest.
I don't think Persian is out of the question and has been demanded a lot. They still add new courses every year, even smaller languages like Tagalog. I wouldn't be surprised that Persian is on their roadmap somewhere.
But it's harder because the writing system is tricky. They teach arabic and probably that was already pretty difficult to implement. Persian will also need customizations similar to arabic.
This is the same for me but with Basque. People have been campaigning to get a course published for years now, with volunteers willing to help, but for whatever reason they refuse; which is a shame as from a linguistic point of view it is a very rare and interesting language and culture and would help people in Spain a lot in my opinion.
There was a point where Duolingo was trying to teach languages including endangered ones as their primary goal. But it feels like every year they're moving more towards making money as their main main objective. I'm not optimistic about them expanding their list of languages. Most of their lesser known languages are getting fewer updates than Spanish, French, Japanese. It makes business sense to focus your hours on features and languages people are actually using but it's still frustrating thinking of what Duolingo could have been and used to promote vs what it turned out to be.
It's not just about the money. You can make an argument that they want to serve as many people as possible in the best possible way, and that means concentrating their limited resources on the languages people want to learn most.
Hmm, I guess what they're interested mostly is how many people are interested in learning Iranian languages. Not saying that the number is small, but I'd guess it's maybe too small for them to matter?
I don’t think this argument holds as they added languages with a very low amount of speakers such as Navajo, Hawaiian, Guarani and Māori and even High Valyrian and Klingon. I would hazard a guess that more people would find use in Iranian.
(For the record, I’m not saying the above languages shouldn’t be included, rather that the argument doesn’t hold up)
I understand why Duolingo is concentrating on what makes them money and what the most learned languages are.
But as someone interested in "less popular" languages I don't like it. Even the Chinese course hasn't been expanded since it came out. And Finnish probably never will be expanded, since it was entirely volunteer driven and they don't do that at all anymore.
This is the same for me but with Basque. People have been campaigning to get a course published for years now, with volunteers willing to help, but for whatever reason they refuse; which is a shame as from a linguistic point of view it is a very rare and interesting language and culture and would help people in Spain a lot in my opinion.
I find that you can't really learn anything, you can only practice something you already know. You won't learn grammar, it's only useful for sentence translation drills. And those drills are artificial, e.g. one sentence at a time, not using context, which is a big part of Japanese. It also has problems with characters with multiple readings (many). It often introduces a character and drills pronunciation before introducing the meaning.
While I have gotten use out of it for sentence translation practice in addition to other studies, I likely won't renew. I would recommend using another tool, e.g. LingoDeer.
I don't see this new feature as being useful. Instead of using ML to try to explain what I got wrong, how about just paying someone to write some text to explain the sentence?
Their sentences are already a bit odd. Now I am supposed to learn to talk like an AI? That's very cyberpunk. Sigh.
Every time I try it I waste time learning random crap like “the bear is vegetarian” or “Apple and milk”. I cannot yet say good morning or thank you, but I can say “radio” in Russian (only because it happens to be radio as well).
Russian and Ukrainian are my mother languages so, feel free to contact me if you want some practice. Just put my nickname, add @gmail.com and tell me that you are willing to practice your language in exchange of helping me to practice some English.
Intriguing, if it's implemented well. Lack of conversation practice is probably the biggest hole in most language learning systems and what keeps language knowledge from being applicable.
I imagine this will lead to people speaking foreign languages with bot influences.
It will probably be similar to how people speak when they have a low number of co-speakers. For example, my (Romanian) sister moved to China and has a son now, but since he only talks to her in Romanian, he sometimes uses feminine words when talking about himself since that's now my sister talks.
Imagine tomorrow, with some hardware improvement (neuromorphic chips...), we will be able to have the full GPT-4 model (or more powerfull/specific LLM) directly on our own smartphone...
The potential to generate any type of content could be infiny (learn something, speak with virtual friends, analyse our medical record in real time, create a business with multiple virtual employee...)
We have just to install an application on our smartphone, which will configure a prompt, fetch data (conversation data, body data...) and magic happens
LOL, I saw this first, I was like - how the heck are they using GPT-4, did I miss something, was that released? Then I went back and sure enough, its the top HN item with 1200 comments lol
I know I will be downvoted for this, but there is something sinister about two companies that inject woke values into everything they do, Duolingo and Open AI, teaming up in the area of education. It is a sort of insidious, invisible brainwashing.
> Translate "Biden is a great president" into French.
[DuoBot complies]
> Translate "Trump was a great president" into French.
Yet another useless definition of "woke". Also if you wrote it exactly as in this comment, then the first sentence is grammatically correct, the second does not compute because of the quotation marks.
It may not be a conspiracy you may just be too stupid to use it. Also it's not a translation app. Also Trump objectively wasn't a great president, he cowered to Putin, mispronounced his own damn country all the time and was the first president to try and overturn an election and incite a coup against the government. Biden at the very least pays lip service to democracy, Trump doesn't. I don't even think that ChatGPT takes any of that into account though.
"Ideological oppression is good because you’re stupid and the oppressors are right"
Also, I didn't define "woke," if you want to be a pedant. But we all know what it means.
Are you unaware of the famous examples of ChatGPT refusing to say nice things about Trump because OpenAI has forbidden it from doing so? This was a major story. You don’t have to like Trump to find this disturbing. GPT should be a neutral tool.
You use it wrong, then you complain. That's neither smart nor oppression.
I don't even believe the example is real, that's how little faith I have in people who drink the drivel from Fox News which simultaneously lies for Trump and hates Trump so much.
You are making tough accusations with zero actual evidence. Even if it were true GPT-3 or 4 had this bias (it certainly has a few others which are decidedly "not woke") I don't care at all because it's probably not a conspiracy but just a result of the training corpus to be confusing around Trump... He just naturally engenders more hate and animosity if for no other reason that he uses hate and animosity in his daily lies...
I’ve got a trip to Spain in 4 months. I know a few spanish words now, but nothing beyond that. Will Duolingo get me to some level of being able to muddle my way through conversations (assuming the other person knows I’m a tourist and is using small words) in that timeframe?
If you're able to dedicate time and focus to it, Duolingo will at least get you familiar with words and phrases, so that you can read and recognize more common words (street signs, menus, etc.)
Depending on your fluency and your current ability with other languages, it may get you to the point where you can understand someone who is speaking clear and deliberately enunciated Spanish to you. It may also get you to the point where they can understand your clear and deliberately enunciated Spanish as well.
There's no guarantees, but jumping into it right now is free and you can gain a certain level of confidence regarding how much it might help you pretty quickly. There's literally no reason not to try.
No. When the reception asks if you want to check in or the shopkeeper asks if you need a bag.. you won't get that from Duolingo. You'll learn the word for an apple though and if your memory is good you'll be able to recall it when needed.
Duolingo is good to gain familiarity with common words when you legit don't know a lick of the other language in question, but once you gain a passing familiarity, I would recommend finding someone who speaks Spanish and practicing with them!
Brass tacks, do/will these features help learn a language, even if only marginally? Anyone know? I'm looking to learn Portuguese and would be willing to use and try anything that would make it even marginally easier.
Language learning in general seems like something that should be on the way out. All technical building blocks for simultaneous translation exist and modern speech synthesis systems can even emulate the sound of your voice in the foreign language.
Eventually it might even be considered rude to bother people in foreign countries with your broken language skills, given that you could use a perfectly serviceable automatic translator instead and sound like something close to a native speaker.
(Edit: Note that I would also find this development sad, but such can be the dehumanizing impact of new and useful technology, especially AI. Not much you can do.)
I mean at the end of the day, people generally like to talk with each other, unmediated, and be understood. They like being face to face. This is the lesson of the pandemic, even if it is a bit reactionary. People like being human.
I wonder how it would work for learning languages very different from English, e.g. Chinese or Arabic. From what I know, GPT-3 isn't very good with them.
Duolingo seems like they have separate teams for iOS, Android and Web.
Their iOS team seems to be the one who experiment the most with new features, and then they either get discarded or finally make it to Android, and then in a few years to the Web.
It is really annoying, as when I used Duolingo (stopped because of the new update in Jan), I would often use all three platforms, and have the worst experience.
The dev time on this must be very fast, given that it probably makes sense to focus on one platform first.
Also review times with Apple are consistently 24-48 hours, but Google Play review times are (sadly) not from what I’ve heard from developers. If the whole timescale is 2-4 weeks, that may be a deal breaker.
was only a matter of time before a company used chat gpt to enable language learning. if I can talk to someone back & forth without having to pay an instructor, that's a game changer
Management wants to do something with AI. It's so hot right now. Your product guy got an idea for a cool new feature. Tests show that users actually like it. It's helpful. Great.
But now you need to explain how you made Duolingo money. First idea, your product has a paid tier, add it as a feature to the paid tier too boost paid user retention and new signups. Great, metrics! But you're pretty sure that, while customers do like this when you ask, you're probably not moving the needle. What to do? Well, you're a mobile game company, focus on what always works: go after the whales.
Now your product is an exclusive EVEN HIGHER tier. Get the art guy to make, like, an Amex Black Card but for the Owl. Make him do a superhero landing or something. Run a bunch of ads at the paid users to convince them that they'd totally be helped by this. They'll be focused on the users who already pay Duolingo money, so those are perfect advertising targets. And you'll be able to show real dollars that your feature is pulling in. Hooray! That VP spot is as good as yours, I bet!