Stonks & Bonds. lol fundamentals, sir this is a Taco Bell

Have you tried GPT advanced voice for translation? It’s pretty clear that there’s no technical barrier to real-time voice-to-voice translation.

I’d be curious to hear how well it knows Filipino.

1 Like

I use the translation and language stuff in chatGPT all the time. It’s very very good. It’s not going to be able to replace me talking to my kid in the same language he talks to his grandma in.

And as others have pointed out. There’s an inherent delay factor to a translation. You can’t know how to start a sentence until you know how it ends, and you may even need the next sentence or two as well. This is not a technical issue. It’s just how translation works.

Additionally a translation never captures the full meaning, it’s always a judgement call to provide a sentence that is as close as possible.

I’m fully convinced that these tools will do wonderful things with translation, and I will use the heck out of them all. But I will continue to learn Filipino.

I spend more time learning tagalog now because of ChatGPT than I did before, and I expect that’s likely to be the trend for most people.

2 Likes

I suspect that translation is likely to be similar to self driving cars

The last 1% of problems is going to be orders of magnitude harder to solve than the first 99%, and the last 0.1% even harder, and so on.

Maybe it gets there eventually, but humans like to be able to communicate across the full 100% and will continue to put in the effort to do that. And if they have a copilot helping them do that, even better.

1 Like

Have you used the new voice mode that came out this week? It’s a game changer for language.

I think simultaneous translation is much more doable than you think. Most sentences can be translated piecemeal, and if the model isn’t sure what to say, it can insert fillers like “um” to paper over the gap. I agree it won’t replace deep personal communication like in multilingual families, but for travel and business, it will be fine.

But I think this is missing the real argument that Duolingo will go to zero. It’s not that people won’t need to learn languages anymore. It’s that they’ll learn using LLMs as teachers and conversation partners rather than gamified apps.

1 Like

I’ll give it a try.

And regarding Duolingo being replaced by AI as a learning tool. That I’m more open to. Although they are leaning in pretty hard to AI tools to try to prevent that.

yep. i’ve off and on’d spanish on duolingo forever. when i got a good streak going i’m in, when i mess it up i forget about it for a while.

i think my issue is i’m old and my memory sucks. without having anyone to converse with, it’s just memorizing nouns and conjugations and whatever.

i can get my point across and “despacio por favor” my way through listening and understanding, but after a month or so, duolingo starts to feel more like homework than anything i’m ever gonna use.

after posting that tweet i actually read through the comments and it does do a good job of getting you involved/invested, but i still think at some point chatgpt will be better at whatever is quantifiable in teaching languages. couldn’t tell you why though, probably bc i’ve been playing with image creation more lately and if it can do that… :joy::joy:

Overheard yesterday by people leaving the Jerry Seinfeld performance.

“Yeah man, we are building data centers 24/7 right now. If they could, these companies would order 100K more data centers to be built immediately, and its almost all for AI.”

(Jerry had mentioned something about AI in his set)

Only a small subset of language users need deep language understanding, like a family. I regularly communicate with clients in China who do not know English. We use a translator (bilingual paralegal or other lawyer) for important stuff. Duolingo is dead imo.

I don’t think it will be dead. It’s a game first, and now they can nearly zero out the cost of content creation while still feeding consumers ads. People will still play their game. But I think the advances in AI give people a lot more low-cost learning options so I see no reason to expect Duo to go on some Nvidia-like run

But who knows, maybe WSB will get on it. Nothing is rational.

I’ve never really encountered someone who uses Duolingo achieve fluency in any language.
Admittedly, my bar for fluency is probably really high. For example, could you take an upper level college class in your non-English language and do close to as well as you could do in English or pretty close. No, then not fluent, imo (I realize this is probably not the standard definition).

Anyway, the point is that no one who learned with Duolingo should probably be making complex medical decisions in that language. They need a proper translator. Not sure what you’re referring to with binding verbal contracts as reasonably important contracts are generally written, but if you’re bargaining for souvenirs at a market, Duolingo should be fine as should an AI earpiece translator.

This is always a hard problem. Right now the best possible solution might be like what they do at the UN where you have a high-level fluency interpreter translating in an earpiece in real tie. However, I really believe that AI should be able to get close to that in my lifetime.

Even that may be annoying, but I cannot imagine a better possible solution that having a fluent human talking in your ear in real time. AI has got to be able to get close to that. Even a human has to think a little bit to be able to do that. It has to be hard AF.

1 Like

AI has zero bearing on my wanting to learn Spanish. I already know enough to get by in 98% of gringo situations. I want to learn it so I can have real conversations with people.

So maybe for travel and business stuff, yes, spending time learning the language might not be worth it. But not so much not for family stuff or being an expat or slow travel in another country.

I’m not sure how duolingo’s users break down among those categories.

1 Like

All the Suzzers in the world cannot save Duolingo. There are too few. If it has use it would be for the rest of the world to learn English for economic opportunities, but I expect there are many cheaper options.

In general, most people do not seek education absent compulsion.

1 Like

If you’ve run the numbers on duolingo’s business model, feel free to share it with the group.

Otherwise I’m going to assume we’re all just talking out of our asses as usual.

1 Like

I think 90% of AI is bullshit, but I see translation, even very good translation, as an exception.

I just received a translation of a half hour proceeding in Chinese. First, AI did transcription of the recording then an AI assisted translator translated the text. It was done by the opponents translator, so I now need to sit down with ours to ensure anything important is accurate.

Meta just showed off this functionality at a presentation last week, I don’t remember if it was part of their current smart glasses or their next gen project version but it’s coming.

ETA: It’s their currently available glasses:

1 Like

I think most people learn a language for the practicality, not the experience of connection. Most people.

1 Like

Everyone I know who talks about duolingo or posts about duolingo (sample size maybe 6) was doing it for enrichment basically. One is Korean-American and wanted to learn it to talk to her grandparents. The rest are trying to learn Spanish mostly just for the sake of doing it. I assume they have a component using it for business, but people just don’t post or talk about it.

Could be tough to admit?
Alsoz maybe im wrong

The vast majority of people who learn a foreign language are doing it neither for family nor to talk to business clients. Real-time translation stuff is exactly like self-driving cars in that people will keep predicting it’s just around the corner but in reality it’s still a decade+ out. These tech companies aren’t in the business of real global solutions. It’s all mostly regulatory arbitrage and cynical value extraction.

2 Likes

I worked in the translation industry for many years. Exited about 10 years ago for personal reasons (I was sick of it). Turned out to be good timing as the group of translators I regularly interacted with are now in a blind panic, and rightfully so, as their careers are about to be made obsolete.

For years, we would laugh at the outputs of Google Translate, comforted by the fact that our jobs would be secure for many years to come as translation was the Holy Grail of AI that could never be achieved.

No one is laughing anymore.

6 Likes