hfhdjdks

Let's say you collect stamps. I think it's a stupid and morally irresponsible way of using your free time (you should be socializing or doing sports or learning something new). Just being honest with my opinion.

This is all ok. The problem is when we start firing people or shunning them for stuff we don't like.

So Chinese LLMs are bad actors, but USA LLMs are the good guys?

I don't see it that way, but I'm sure from an American perspective that how it seems.

This sounds like a very negative comment and with many generalizations. Please consider thinking about what your comments bring to the discussion

>Since this theorem is true for the standard model of the natural numbers but not provable

I've always found the provable vs true comparison confusing. How can we say the statement is true under the standard model if we cannot prove it? I understand that it could be true, but how do we know it? If it's proven with second order arithmetic, then this implies it is true under the standard model too?

Or are there statements true independently of the axiomatic system you use to prove them? (Apologies if this is too off-topic)

The point is not whether it can be modeled (it can), but whether you can get a closed form solution for it (you can't...or we don't know how for now)

I see a lot of people asking "what year it is" or some other question that tests knowledge lf current events. To me it sounds like a misunderstanding of what an LLM + retrieval augmentation does.

If the LLM is only fed the HR documents that have multiple dates, why would you expect it to answer that correctly? Why is it relevant that it answered it correctly.

"are you a human?" "No, I'm a model trained by OpenAI"

Nearly fooled me

Not in India, but I've seen parents here sit down at a restaurant table and shove a cell phone in front of a toddler the second they sit down. They don't allow the kid to even get bored, try to do other stuff or interact with the kid. In my experience, a 1-2-3 year old kid doesn't have the "cellphone have cartoons, gimme!" unless they've been educated to expect that.

Does that seem like a bad habit to impose on your kids + you not engaging with them? It does to me. Am I gonna call (the equivalent to) child services? No. Do I make other probably equally bad choices for my kids? Quite likely.

No need to get defensive.

Or maybe there's no secret sauce for intelligence and if the system / organism can display all that functionality then should just say it's intelligent.

I don't have a strong opinion either way, but I'm not convinced by the "secret sauce" / internal monologue school of intelligence.

If we want to be pragmatic, we should just think about smart tests and just assume it's intelligent if the system passes those tests. It's what we do with other people (I don't really know if they feel inside like I do, but given they are the same biological beings, it sounds quite likely)

Faang salary only exists on the US. Even the richest countries in Europe don't have those salaries for programmers (maybe if you work on finance or some niche subject).

In this part of the world, we're more worried about what the US government can do to us than what the Chinese can do.

In other words, for many people in the world that wouldn't be an issue, it's already in the hands of a country that had invaded or intervened far more countries than China in the past few decades

Not an argentinean, but I don't think that's the right way to look at it.

You can exchange dollars in any country and in many you can directly pay in stores if the storekeeper is flexible. You wouldn't call that a parallel dollar economy.

The only additional point in Argentina is thay they have an official exchange rate that is way way different from the true exchange rate (ie the one that arises from demand and supply). That's why you can exchange dollars through the official channels and get, say, 500 pesos, or you can exchange from an unofficial one and get 1000 pesos

The language is not dumb. The writing system that people chose is...not that great (and you can call it dumb if you like). There was a choice between respecting phonetics or maximizing some other objective (respecting etymology or being consistent with older pronunciation or whatever). Being based on a latin script maybe wasn't ideal, but languages like swedish or german did a better job and invented letter when needed.

If you take the distribution of writing systems, I'm not sure English is particularly bad (the modern version has spaces and punctuation! Not bad, right?). It's not great either.

>English does not have a regulatory body like Spanish or German have.

Spanish does have a regulatory agency (RAE) that we choose to follow, but AFAIK, they don't say anything about pronunciation. Spanish from Spain, Mexico, Colombia all have different pronunciations and not one is more official than the other. There's a certain sense of what's "normal" or nearer the center of gravity for most speakers, but I think that's true for English too (new zealand or south african English is less "standard" than a Midwestern accent or BBC English)

The didn't recently have a president looking to undermine the very fabric of democracy. That already makes it better than the USA

>but if I had to choose a country to get ahead in life with the most opportunities possible as an immigrant

The USA is pretty bad (compared to most rich / developed countries) in general at giving opportunities to the broader population. Check the studies on intergenerational mobility to see that the USA ranks pretty badly there. It's not surprising to me given the inequality of the education system and how segregated it is (in terms of social class, parent educational level, etc).

You bring an interesting point though: "as an immigrant". Maybe the USA IS pretty good at giving opportunities to certain kind of immigrants (even if it isn't for the broader population). I haven't seen any study about it, but I'm sure there are analysis.

The busting and splitting up is a hard tool, but anti-trust has many other tools at their disposal, including maintaining status quo (i.e. refusing the merge of two companies). Disclaimer: I don't work in the US, so I'm not sure if anti-trust there is heavily biased there to this heavy duty tool, but I do know they have plenty of other tools

Sounds like a two-sided market, where a company gets their value from being able to connect one side (the artists) with the other one (the venues).

What I find strange is that people aren't usually reluctant to sign exclusivity agreements because it's very limiting. How did ticketmaster get so big to begin with to force / convince people of signing exclusivity clauses?

My lazy guide:

This didn't make my daughter 100% bilingual, but she's up there.

We live in a spanish speaking country. Netflix (where she sees all her cartoons and movies) has always been set in English. She's 7 now and has never asked for it to be switched to Spanish.

Additionally, I read to her almost every day (before bed, sometimes during the day too) and 90% of the books were in English. It would have been tough if, for some reason, she didn't liked the English books, but since I started before she could even speak I think it came naturally to her.

I used to speak in English with her (and she would reply), but at some point gave up.

Her listening skills and reading skills are amazing (and on par with her Spanish), but her speaking ability is not quite bilingual.

Also 2 == 2. You're not making an argument, you're just saying that LLMs aren't human.

This comment isn't saying anything. It says "LLMs statistically choose the next token" and, because of that, it can't be anything else.

We know how the LLMs were trained, so re-stating it doesn't help in anything. The point is that after the LLMs are trained they behave in certain ways and it can be helpful to say something about how it behaves.

For example, we can talk about how a linear regression can or cannot capture the causal effect of X1 over Y. "It's not capturing the causal effect. It's just minimizing the squared error" is an unhelpful statement.

The GP is analyzing some dimensions to figure out if the success of Baldur's Gate is surprising or not. He/she finds that, given the limited dimensions we can consider, the success is surprising.

I read your reply as saying "no, it's not possible to analyze why games fails / succeed, because they're all different". I feel that's usually unhelpful: assuming we can't explain things because they're all idiosyncratic is usually not productive. It's more productive if you, for example, point to something extra that is missing.

This is an old question in economics, but undoubtedly relevant now:

"Is this new technology change different or similar to other technology changes?"

In other technology changes, there are hundreds of jobs that disappear, but hundreds that appears...and, as far as I know, we haven't observed an increase in joblessness in the last, say 30-50 years. We went through the industrial revolution, the invention of personal computers and a big etc and it didn't happen.

Is this one different? I don't have an answer, but I don't think the answer is obvious.

I grew up used to reading subtitles. Listening to the original voices and languages is super interesting to me. Besides some very specific cases (I can't look at the screen all the time) I don't have any use for dubbing.

You're correct. Parent comment says something true (some / many characters don't have a representation in most fonts), but the inference it makes is wrong (this is a bottleneck for communication)

Does this come from actual experience trying to communicate in chinese or japanese? At least for chinese (but I'm pretty sure it applies to japanese too), what you're saying doesn't ring as relevant (i.e. it is true that some characters don't have a graphic representation in many fonts, but it doesn't cause problems for communication / creation of new words)

Plenty of fonts can represent a vast amount of characters (search around, but even free fonts can have >50k), much more than even highly literate people could ever use (a very educated person might know 8k characters, but would use much less actively).

Every new release of Unicode releases new characters that are slowly incorporated into fonts...but for many years these are on the truly long tail of characters, nothing that any common person would use. There might be a few exceptions with new characters (the new japanese emperor characters comes to mind), but the thing to bear in mind is that new words don't necessarily involve new characters. A big chunk of nouns in chinese are composed of multiple characters, so making a new combination can generate a new word (电视 is television...electric + vision). This is probably even easier in japanese were they can use katakana to spell an English word and create a new japanese word.

Are you american by any chance?

Over here people speak multiple languages. I doubt we'll run out of people that speak multiple languages just because there's a language model that can do great translations.

I take it that your expectation is that a woman takes more care of the house?

The really good movies are immune to spoilers and you can watch them many times.

But there are also plenty of ok movies that are fun, but mot rewatchable. I usually enjoy a movie more if I watch it without knowing much and getting surprised.

It's not what you were asking for (class that can't be subclassed), but `typing` has an `assert_never` to check exhaustiveness:

https://typing.readthedocs.io/en/latest/source/unreachable.h...