MiroF

> these bubbles only represent a small fraction of the totality of these places,

You are substantially under-estimating the size of the urban population of these countries if you think that the major cities only represent a "small fraction."

> That's why they are willing to compete for competent engineers.

The skyrocketing wages for engineers are because until recently, tech companies were illegally conspiring to not compete for competent engineers.

> 10x developers

If they're causing code quality slide, they're not a "10x developer."

> They WANT to work under heavy regulation

Well good, I hope they'll be appreciating my no vote on 22.

Ah yes, the average Uber customer as the "little man."

The well-being of others in the society I live in is of concern to me.

> Seems like the fact that there is constantly surge pricing and long waiting time means that there are too few Uber drivers, even in "low quality" cars, to meet demand. Your complaints sound like the old joke about prison food: it tastes terrible and the portions are too small.

What an odd thought leap! It seems like your comparison to the prisoner would apply to anyone who thought the price of something was too high for what you were getting, so doesn't actually seem like a compelling point.

Uhhh.. I just translated the exact same sentence from wikipedia in Google translate and got:

"The juvenile protection procedure aims to deter and rehabilitate juvenile delinquents, and as a process up to the decision, "send and notify the facts of delinquency to the family court-family court investigator (hereinafter abbreviated as" investigator "). It is customary to go through the flow of "investigation by, etc.-a referee based on the investigation results-determine protective measures or protective measures as necessary"."

Not incredible (especially when it starts listing the process), but not as god awful as whatever you got. I'm curious if you omitted some part of the sentence when pasting in.

It's also worth noting that this is a pretty out-of-domain legal Japanese sentence.

> This is due to the fact that internally "пружи́на" is translated to English first, resulting in "spring"

Do you have a source that that is actually what is happening behind the scenes? I work in this field and would be surprised if that is what Google is doing. My guess is this phenomenon is more due to a model that was trained on an Russian->English corpora first and then trained on Russian->German due to low resource, or something like that.

Someone like Yandex is going to have better translations because (shocker) they have larger russian corpora than google.

You're absolutely right with your impression of GPT/these models - they will generally generate plausible text.

And it is also worth noting (as you've sort of mentioned) that appearance of fluency can bias translators who are instructed to just evaluate accuracy, or to evaluate accuracy and fluency separately. There is ongoing research into that problem.

Yeah, there's lots of phrases that can be unexpected when you first run into them - so I can get how you'd find it confusing.

I don't just mean 'rich countries'/'poor countries.' First, because that is not necessarily as simply true anymore - the economic relations are much more complicated. Second, that is a value judgement about how countries ought to develop.

You can read up more about it https://en.wikipedia.org/wiki/Global_South

> Languages that aren't related are not, regardless of whether you have a large corpus to work with

The size of the corpus makes a huge difference. I don't see how can both be true that a. there is this massive failure of translation engines on these language pairs,

b. human translators were unable to differentiate between human translated chinese-english sentences and machine translated sentences.

It's not incredible - but it's also not awful or unusable.

"the disfluent but more likely accurate results"

-> Human evaluation does not agree that PBSMT is more accurate than NMT. Sure, the fluency/"sounds good" gains from NMT are much larger than the accuracy gains, but evaluations I've seen put NMT ahead on both metrics.

> It's literally unusable

It's not "literally unusable." Even for language pairs with poor translation, it considerably saves on manual translation time if you can have a translator look over sets of already translated (but lower-quality) sentences and correct those with errors.

Plus, there are actual studies out there comparing human translations to machine translations in a double-blinded evaluation by translators and the gap is not as high as your comment suggests.

For chinese-english (the other direction), bilingual evaluators have had difficulty differentiating between the machine translation and the human translation.

> english <-> east Asian languages

I'm curious what the domain is and what languages you're talking about, it's hard for me to evaluate your claims otherwise. I would buy it for English -> Tibetan for instance.

Mm... I think you'd be hard pressed at this point to find any advantage that traditional methods have over deep methods.

Maybe this was true 4-6 years ago, but no longer.

> more distant languages,

low resource translation is definitely still bad, you're right.

> near perfect translation just by putting each word through a dictionary, and the neural networks can disambiguate things and smooth over the problems in the output

Funnily enough, this is not a terrible explanation of how the current sort of approaches to low-resource MT work.

I don't think I'm emphatically stating it, but I can see how you would read it that way. I added a clarifying reply, but it is too late for the comment to be edited.

To be clear - I'm not stating that there aren't legacy preferences or that the increased admit rate is solely or even majority due to the high test scores - it's not.

A legacy with a 1600 has a substantially higher chance, due to their parents alma mater, than a non-legacy with a 1600.

You're the first person to use the word "peasants" to refer to non-legacy applicants and I in no way think there is some sort of "intellectual" superiority or elite class of people in the US (well, economic elite - but not intellectual).

All I wanted to clarify is that these legacy processes only really come into play once you have already hit some minimum threshold of merit, and there are different backdoors for super rich/connected applicants that require no such threshold.

Tests are not the same thing as intellect - they are often structured in a way that is unfair to people with low incomes (for instance: the more times you pay to take a test, you can then select the top scores). All of these are pathways that would still exist without legacy.

You're construing my comment as taking a stance that it wasn't - but in case it wasn't clear now, legacy admissions are unequivocally bad. Privileging class status in admissions is obviously bad.

I'll try to find a source for that, it's somewhere in the SFFA trial documents but I'll need to go back and look.

But to be clear: I'm not trying to claim that there is no legacy preference and that it can be all explained by test scores. It is definitely a big factor.

The only two points I was making is that 1. traditional "legacy" admissions is not necessarily the route that a lot of students without any merit are getting into these top schools - they often have separate pathways for donors and super elite individuals, like Harvard's "z-list."

2. Legacy admissions are because schools know that if they admit more legacy, they will get more donations - but it also has to do with calculations about selectivity. Harvard wants to maximize the chances that the people it admits will choose Harvard so that they don't have to admit more people to fill the class thus raising the admission rate.

That's not what a legacy admit is defined to be.

I'm not claiming it's a deterministic process, but it is worth noting that Gtown is the best in DC

This podcast is presented as if testing is an American phenomenon - but it hardly seems like it. If anything, the extreme parental backlash against testing seems more of an American phenomenon.

I've had the opportunity to apply for a few british and french things in school/academia and both required that I take a test - and were actually more test-driven than the US appeared to be, with explicit harsh cutoffs based on your test score's z-score.

Before even moving on in the Oxford application process for physics, for instance, you have to first pass a test putting yourself 1.5 z-scores over the average, then pass a series of interviews where you are timed on your ability to ask questions to you posed in person. And that's already after your A-level exams in highschool or whatever.

> There are many poor people in Thailand but many very, very wealthy people and you might not be able to recognise which is which. The wealth divide is very big.

Of course - this is true of most countries in the global south since the 80/90s.

This seems like just classic poorer country trying to milk money out of rich westerner (well minus the floor dropping out that sounds awful)

PHP is literally the only language banned in some major tech companies because of how dysfunctional it is.

That is for undergrad, and it's worth noting that the average test scores of an admitted legacy student > the average score for all admitted students.

Yield calculations have a lot to do with legacy admission rates.

JS is losing out to TS long term.

Python is getting Mypy.

PHP (and Ruby) are dying much more rapidly than many of their peers of similar ages.

> missing features (like unions) when working with other languages.

Which languages are missing unions? Off the top of my head, they exist in C/C++ and mypy python.