LLMs dont have the same usecase as encoder only models. Lets assume you have around million keywords and you want to find the most similar to a keyword that the user input.
In pre-processing you would have calculated the vector encoding of all the million keywords before hand and now with the keyword the user input, you calculate the vector and then find the most similar vectors
LLM is used by end user, encoders are used by devs in app to search/retrieve text.
It takes around 20+ years to go from H1b to permanent visa/green card. In the meantime your kids born in US have grown up, graduated, you have a house and everything could be yanked at the border when you are travelling.
Meanwhile vast majority of them pay into taxes and social security and leave the US and never see a dime of that money.
Immigrants are the easiest group to exploit by everyone because they have no voice and are vilified by vast majority of the people include the so called intellectuals in here.
The byzantine US immigration system absolutely is an impediment to people coming and staying here, and in my (admittedly anecdotal) estimation is a major competitive disadvantage, and a big part of the reason the UK, EU, Canada and China are making progress towards becoming tech hubs.
Canada is not making any sort of progress towards becoming a tech hub. Canadian engineers' dream is to work for a US company. Canadian investment landscape is just sad, but that's a different conversation all together.
Those particular cases have benefited by the shortcomings of the USA actually. I know some big tech companies send workers who weren't able to secure US immigration specifically to offices in Canada, the UK or the EU. For example Meta and Google [1][2].
One can expect the company then grows an interest in developing full engineering teams in these sites. One can also expect some people might simply decide to not come back to the USA.
With the general rise of China's tech scene, recently there's been a trend by which the USA doesn't retain Chinese international students and they instead opt to return home. One has to imagine the very, very long immigration process they have to go to has to do with this [4].
Outcomes aren't binary. For any marginal increase in immigration difficulty for skilled tech workers, there is a marginal decrease in US tech competitiveness relative to other countries.
20 years only if you're born in India married to someone born in India. Not great either way though, but it's really affecting the Indian community because of their particular norms.
If your spouse is a US citizen or permanent resident on their own, great. But if you're on H1-B and your spouse is on H-4, I don't think their country of birth makes a difference?
If you're both on H1-B, then sure, having a different country of birth can help.
It does make a difference. When filing for adjustment of status, you can request USCIS to consider both you and your spouse as chargable to your spouse's country of birth, and therefore be placed in a more favourable GC queue. This is called cross-chargeability [1].
Because of this, the "100-year green card queue" problem only really applies for a couple who are both born in India/China, with kids who are not born in the US. If even one child was born in the US, they would be able to sponsor both parents for an immediate green card when they turn 21 years old. In the meantime, the H1-B beneficiary can extend their visa indefinitely and port their approved I-140 whenever they switch jobs, with a 6-month grace period. The spouse also has full working rights.
21 years is a long time, but while working, both parents will accumulate social security credits and will be eligible to recieve benefits upon retirement (if they've secured a green card by then).
I’m sure this is no consolation, but as a born-and-raised citizen who has paid into social security for 15 years now, I have serious doubts about seeing a positive return on those taxes myself.
>I’m sure this is no consolation, but as a born-and-raised citizen who has paid into social security for 15 years now, I have serious doubts about seeing a positive return on those taxes myself.
Look at the bright side though: You get a chance to get conscripted for a war against Iran/Russia/China and also get to blow up windowless mudhouses in the desert to protect democracy and freedom back in the states.
If only they just used the current 'investors' money to pay the retirees. They actually use the social security taxes to pay for "whatever" and hope they can come up with the rest when they need it.
That's not how social security works. You're not supposed to get a positive return. You directly pay a basic income to retired people (minus administration costs). When you retire, workers pay a basic income to you.
The issue is that, for me and anyone else who reaches retirement age after 2034, only about 80% of that basic income will be available. For reasons I'm not super clear on, this idea tends to get coded as a conspiracy theory in many circles, despite being uncontroversially true and widely reported on.
That's a perennial Boogeyman. Policymakers have a wide array of tweaks they could make (from adjusting the cap to adjusting retirement age) at any time that could push that out by another century. https://www.epi.org/blog/a-record-share-of-earnings-was-not-...
Asking for a positive return on social security is like asking for a positive return on welfare. The positive return comes from not having so many homeless old people all over the country. It's not a personal investment vehicle.
It could be that OP expects Social Security to be kaput by the time he gets to be old.
Looking at the population graph, that’s a valid concern. There’s a ton of boomers and a ton of millenials, but very few babies to pay for our retirement.
(This phenomenon could invalidate even individual stock investment retirement plans as well. We need a future generation of workers, investors, entrepreneurs, consumers).
You are confusing what he is calling for. He is not advocating for people who are already citizens to lose their citizenship. He is saying, going forward, people who are born here will not automatically be given citizenship.
I read through it and even asked chatgpt for summary and it looks like "passport is now required" and "one beneficiary one draw" that is if you put in multiple petitions it will only consider you once.
I thought Elon was talking nonsense when he mentions frivolous government rules but reading these h1b changes makes me question my own sanity about the government "rules" which they aptly named it as "Final rule" (wtf?).
Ah yes the classic "huwhite people and muh racism" gambit.
It's tired. H1Bs are gamed to the point of uselessness. Most companies internally post H1B job offerings so people are aware. I've yet to see one with a competitive salary. They are used to source cheaper labor and avoid paying actual Americans the fair wage they deserve. The last 15-20 years of tech has slowly seen the InfoSys-ization of the tech economy. I work with more contractors from Mexico, India, and Eastern Europe, and more H1Bs from India than literally anyone else. On my team I can count the number of Americans on one hand.
The program should be extremely limited. I am a fan of charging 2-3x the normal tax rate for H1Bs so companies have to actually justify hiring "talent you can't find in America". There are 300,000 unemployed tech workers. I find it hard to believe none fit the bill. Just that most won't take a 60% haircut for more work.
I have been on H1B forever now, and my salaries have been more or on par with the role. I tend to agree there is a lot of H1B misuse, especially by large Indian consulting firms. This needs to be curtailed.
But, there may be 300,000 unemployed tech workers. While I also find it hard to believe none fit the bill, I believe most don't. So many are out of random bootcamps, self proclaimed programmers who can't solve fizzbuzz. I also have not seen any H1B in my career that is good and willing to take a 60% haircut. In my own company, they are the highest paid and are grumpy we are not paying more. They are all really good engineers too. Heck, when I was looking to move to the US, I refused tons of low paying jobs. When we opened up backend programming jobs, only a handful American citizens even applied. We hired one of them, while we needed 4. The rest didn't make it through the interview process. We also rejected tons of H1bs because they didnt make it through the process. Same salary range offered to H1Bs. And we are a fully remote. So I wonder where are these 300,000 unemployed tech workers.
Cut the fraud and it automatically becomes a decent program. Now, if one is entirely against the program of attracting foreign talent, thats a different discussion.
Well, the way program exists now, it's utilized by two kinds of companies:
1) Someone like Verizon that uses it for cheap labor
2) Someone like Netflix that wants to hire good engineers
The way the program works now (before those changes?), it's much easier for group 1 to fill its positions via staffing agencies overseas. That's true even if a company from group 2 already know who they want to hire, since it's a lottery system.
Would be easier if this were two different visas (or program got revamped in a way that it actually works as it's sold to public), but we can't have "Cheap Human Labor Visa" for various reasons.
every problem has a solution except in America where what we THINK is a problem (and discuss ad naseum on HM) is there by design. Group 1’s lobbyist are paying A LOT more than Group 2 - hence they get the most benefit out of the program. it’ll be interesting to see next four years, I suspect the program will at minimum triple
I have a hard time hiring an American too, especially for backend jobs. But thats because they simply don't apply to the open positions we have. We don't disclose salary upfront, so the argument that "you pay less thats why" doesn't hold. We just don't get those resumes - through recruiters, direct channels, LinkedIn - even when we said we prefer citizens (due to legal costs).
The American people get to decide who they want to allow in and under what conditions. If the American people decide that they should get compensated more than non-American citizens for a role falling under American jurisdiction, they can do that. And other nationalities can retaliate or pound sand, but that's it.
Other countries are free to compete, nobody is arguing otherwise, but it is explicitly the right of any country to determine who is allowed to compete within their nation.
What people want or deserve is irrelevant. If you live elsewhere and feel you deserve more, then that's not America's problem.
That was the context. Your post before this changed it to "want". I was responding to that. Nothing to do with non-americans feeling they deserve more either. It was about why americans feel they deserve
You may characterize it that way, and invite some pretty reasonable animosity; but if you do, then want!=deserve regarding the salary of foreigners, either.
I was paid between 400 and 600k a year while on an H1B.
> I am a fan of charging 2-3x the normal tax rate for H1Bs so companies have to actually justify hiring "talent you can't find in America".
This is extraordinarily racist if you spend more than 5 seconds thinking about it, and honestly you should question every one of your choices that have led to this point. It is time for you to re-examine your entire worldview.
Why write like this? It's antagonistic and pompous. I really don't like making light of racism by leaning on the "stop calling everything racist" trope response, but this is pretty extreme. I have no idea what's in the parent's heart, but you only need to give them an ounce of benefit-of-the-doubt to believe that the quoted sentence comes from a place of simple favor for one's own fellow citizens, and not petty racism. And on HN, you're supposed to be giving even more than one ounce of benefit-of-the-doubt.
To be clear, the actual proposal being made is "I am a fan of charging 2-3x the normal tax rate for H1Bs".
One interpretation is that workers should pay 2-3x the income tax, massively depressing net wages for people on visas.
Another interpretation is that employers should pay 2-3x the payroll tax (I guess Social Security and Medicare in the US?) which again means that (not immediately due to nominal wage rigidity, but over time) visa worker wages will be depressed. In any case, visa workers pay into social security, but will not be able to claim benefits unless they become green card holders.
There are already substantial fees employers have to pay, which already depress wages. The proposals suggest making it worse. There is no real thought behind them, no research, no data. Just pure naked nativism: workers must be punished even more than they are right now for daring to immigrate.
It is, in other words, extraordinarily racist. And if someone, through whatever life experiences, has come to believe that this is the way forward, then they absolutely should revisit their worldviews.
---
Neither of these come anywhere close to addressing the actual problem, which is that it isn't the case that workers on visas have the same labor rights as everyone else. Workers on visas are preferentially hired by some firms because they will silently deal with abusive bosses, long work hours and sexual harassment. Giving everyone full labor rights addresses this issue completely.
Do you want H1B worker wages to be depressed, or do you not? Do you care about your fellow workers being sexually harassed, or do you not?
> I have no idea what's in the parent's heart
I don't, either, but structural racism is a million times worse than some rando shouting a slur at me.
> This is extraordinarily racist if you spend more than 5 seconds thinking about it, and honestly you should question every one of your choices that have led to this point. It is time for you to re-examine your entire worldview.
No. Americans should look out for Americans first. This isn't "racism". It could be interpreted as "nationalism" but if Americans don't look out for Americans first - what's the point of even having a country or a flag? I've spent a lot of time thinking about it. Why shouldn't we make companies pay more if that foreign talent is really needed? It should be in desperation that you reach beyond your own countrymen to find what you need.
Second, I had no idea "Americans" were a race. Maybe it's you that should seek some help.
50 million is chump change for Facebook. In a fair world, we would have competition against facebook and there would be choices for consumers. In this real world we are stuck with the bottom feeders like facebook which just works with impunity.
Not only did facebook lose all the data, it disabled customer support chat in Australia for all services including advertising.
Our company fb account (for travelers) was disabled due to "Suspicious activity" and then permanently deleted. But they conveniently left the ad account on which we can neither login nor disable. If we do a charge back, Facebook will instaban the instagram account as well and in process we could lose our 16 year account.
In fairness facebook has many competitors and it's possible to create a clone for a small sum. That doesn't include the scaling components which are world class. But a profile some photos and videos has been done successfully by vk for example.
How to you get everyone to signup? This is how facebook did it. Start in top universities and only allow people from there to sign up. Repeat at other universities. Open it up to all .edus. Get press coverage about this site many are talking about but only college kids can use. Open it up to everyone but setup cities regions where content is shared. Introduce a tool that reads your hotmail and emails everyone to join. Add games. Groups. Kill games. Promote flame bait. Replace content with ads.
The facebook rollout was magical and hard to replicate. Having people within the same colleges on a public/private network created communities. Places where parents/relatives/past school mates wouldn't see that content. Lost some magic with cities but it still created communities. Killing games started the decline. Everyone you know from your grandparents to your little sister joining killed the original feel. Politics killed conversations.
Facebook had many paths to choose and they ended up here. If you started a similiar journey you could make different choices and end up somewhere else.
The “you are the product” cliché might be the most underrated phrase of the last decade or two.
What Facebook did was not about some genius move where they got university students to sign up; it was about making it seem as if your customers are not who they actually are, and by exploiting the market contrary to how it’s supposed to work. Facebook is not the only one who did it, but they are perhaps the first ones who did it for social media and at such scale.
The bright idea of the free market—customers pay for things they get value from, which helps more of these things to exist—is turned on its head and inside out when the product customers pay for is people paying for things while those people are misled into thinking they are being provided a “free product”.
Any value the end user gets from the system is purely incidental. If it is harmful to the end user long-term, that is orthogonal to the bottom line. The company is not incentivized to provide the end user any value, the company is not interested in helping the end user succeed and prosper—rather, the company is incentivized to not let the end user leave in bulk (that would hurt the actual paying customers, that is the advertisers, and the bottom line) and to keep their eyes on the non-product for as long as possible (if riling them up by algorithmically chosen triggering posts works for that purpose, great).
This fosters monopolies, infects capitalism, and when it becomes the default engine of human interaction it infects society.
I have an account like that that simply exists but can never interact. Support requests (even when I figured out how to make them) do nothing. They just say "It will be fixed within the week"
Neither is "White". An Australian is completely 180 opposite of Canadian in almost everything and there is almost nothing common except monarchy and English language.
Are Amish white? People from Turkey? Sicilians? Argentinians?
I find the whole concept of "race" perplexing and catering to the lowest common denominator.
You could make the same point about the actual color “white”. You can point to various items of different colors and ask “is this thing actually white?”. You’ll have some people agreeing on most of these things, and disagree on some of these things. Would it make the whole concept of “color” meaningless and useless? Hardly.
If you had some people that were from India, some from Japan, some Siberian Russia, then actually I could see the answers being significantly different
You are missing my point: Asians - vietnamese, chinese, japanese, naga indians, have very little in common from educational/ecnonomical/social point of view.
Same with "whites" (whatever the definition is and for the life of me I just dont understand).
An Australian moving to Canada has a culture shock that is inversely similar to a Japanese going to Paris.
Americans are weirdly still clinging on to the race concept
There are common physical traits that form a family resemblance.
There’s also skull shape and nose length. But what’s the point of binning by these traits? US usual racial division is nonsensical (and I’d say offensive if I was american) and no one dares to question it. You literally blow the question of genetical origin to 100x size compared to the rest of the world. And yet don’t even do it in a way that could make sense instead of being a phenomenon similar to racism but by ignorance and meaningless overgrouping. And then avoid talking about it cause it’s a taboo. It is uniquely American.
>It’s probably the place where discussing race has the least social tolerance.
No. the rest of the world doesnt discuss it because it makes the least sense. I gave you examples above. It doesnt make sense to put an Aussie and Canuck in same bracket, neither does it make sense to put a Thai and Japanese in same bracket for whatever data point you want.
Even within India a sikh has nothing in common with naga has nothing in common with Tamils.
Race is an absurd abstract. Why not hair color? or eye color? or eye shape?
to show what I mean. See those words in the context of the sentences in which they occur, and it can be seen that you are not giving any evidence for your claims either.
Do your own research, if my comment is so important to you.
Also this is HN, a forum, not a court of law. Tons of other users on here, regularly and casually make comments which may seem false to others, without giving evidence for their statements.
>Notice that a charged label on a Wikipedia page is enough for you, a skeptical person, to make an absolute conclusion.
Don't try to mislead, by using words like "charged label". The conclusion is clearly made by the Wikipedia page, not by me. I merely quoted it. Anyone who doubts that can go read it first, before making "absolute conclusions".
>Notice that a charged label on a Wikipedia page is enough for you, a skeptical person, to make an absolute conclusion.
Notice what charged label on the page [1]? By common sense logic of conversation, if you considered that I was using a "charged label" (whatever the heck that means) in my argument, the onus was on you to, at a minimum, mention that label, which you clearly did not do, although I think I can guess which one you mean.
>If it’s so obviously false, can you share the landmark study or experiment that disproved it?
How about you first sharing "the landmark study or experiment" that proves it?
And I wonder if you read the whole article, seeing that the word "discredited" (referring to phrenology) appears at least 5 or 6 times in the article, in many cases with citations.
>All that’s true, but you can still identify them trivially. There are common physical traits that form a family resemblance.
That's exactly what I implied by my above comment ( https://news.ycombinator.com/item?id=42383753 ), which means the same as this part of yours: (i.e. you can still identify them trivially. There are common physical traits that form a family resemblance.).
But mine was referring to the differences between Indians on the one hand, and East Asians (and some others) on the other hand.
Here, the ”East Asians” includes Chinese, Taiwanese, Japanese and Koreans, among others, and the "some others" also includes Tibetans, Nepalese, Bhutanese and Sikkimese, and Burmese, and people of the North Eastern states of India, such as Assam, Meghalaya, Manipur, Nagaland, and Tripura.
All of these above people can easily be distinguished from (the rest of) Indians by their narrow eyes.
I don't know what is so difficult to understand about this. Nor is it discriminatory. It is simply stating obvious facts, that anyone can visually ascertain for themselves.
It is the people who behave in a discriminatory way towards anyone, based on where they come from, or their ethnicity, who are the real racists, not those who talk in an innocuous way about distinguishing visual characteristics of various categories of people.
>Yeah, I’m not buying that race is a uniquely American construction. It’s probably the place where discussing race has the least social tolerance.
(
>American
Say the word right, dude. USian, not American. :)
)
Discussing race may have the least social tolerance in USA (or not), but when it comes to actual behaviour, i.e. racism?
And I do not mean to single out the USA. There is plenty of racism in Europe and Asia too. I have myself experienced it in both places, and have heard of plenty of anecdotes about it from people whom I trust.
There's probably high genetic kinship (Britain) for Canada and Australia amongst 'white' people, the US too has a large body of genetic kinship with Britain, it's definitely not your 180.0.
>The US is upping the ante in its information war with China, with the House of Representatives backing a bill to spend US$1.6 billion within five years to promote anti-China propaganda, according to observers.
Sarcasm like this will warrant a spastic response from paul graham, cuban and his dick cheney thank-you-for-your-service hordes about how america is exceptional and lord himself created the country with his bare hands ...
In pre-processing you would have calculated the vector encoding of all the million keywords before hand and now with the keyword the user input, you calculate the vector and then find the most similar vectors
LLM is used by end user, encoders are used by devs in app to search/retrieve text.