What impress me concerning chinese LLMs is not exactly their performance (though they are, indeed, impressive), but the variety of labs producing frontier models.
What is the main difference between a "shelf" and a shareable folder?
For what I understood, it is mainly that you can manually sort the entries on the shelf?
One of those instances of AI-generated code replicating another library's API, using a more low-level language for the core implementation.
I love the idea of using new low-level languages and making libraries faster. But I wonder if there is a better way to recognize and value the hard work that the original designers of libraries like FastAPI, Zod, Pydantic, have invested in to make those libraries.
Without the API design of those libraries, this turboAPI and dhi would make no sense.
I was asking earlier if DHI has tests - seems they do. As does the FastAPI refactor.
Would be nice though if there was an exact test suite that ran against both that could demonstrate parity. It would be onerous to try and compare both library’s test to see which one is more comprehensive (believable) - by default I’d expect the original to be the most believable (duh) mostly because tests represent historical cases that can up and broke it.
It's not a tautology because it's not guaranteed. There are plenty of plausible sounding claims that fail to be true. That's why science is needed: to provide _empirical_ evidence for/against a claim.
Not to be an "uhm actually" guy but this goes into a lot of interesting philosophy in the first half of the 20th century. You would probably agree that "a fish is a fish" is a tautology, but for more complicated things it gets murkier and murkier. Separating out what are the tautologies from not was a big effort. Then Quine came along, and a big portion of people migrated away from the distinction
I dabble in "um actually"s myself (especially given that my original comment was one), so no worries :)
I don't disagree with your comment exactly. But I primarily wanted to push back on a common response to scientific works. Something to the effect of "Well obviously, everyone knew that!".
Except they didn't because they (presumably) didn't actually investigate. And even after the science, they still don't _know it_ know it. But post-scientific inquiry, they have a much stronger claim to the knowledge than they did before. So the type of dismissal in the root comment is seriously missing the point.
I think culture and education play much bigger roles than anything else, all the sources I find show Germany and France having similar level of corruption (on top of being geographically and economically close) but completely different level of "social trust".
China's pretty corrupt politically but the social trust is quite high, the highest outside of northern europe as far as I can tell
The corruption numbers break down into: (1) They didn't ask the question in China, (2) They asked somebody if they paid a bribe or if taking a bribe is every justifiable, and (3) "Expert estimates of the extent to which the executive, legislative, judiciary, and bureaucracy engage in bribery and theft, and the making and implementation of laws are susceptible to corruption"
For (2) China doesn't look too different from the U.S., for (3) experts think it has gotten much worse since the time of Mao but I'd say China is on the honest side of the "global South".
Note that lay perceptions of corruption are widespread in the US
> though unlike India I think very few Americans have paid a bribe to a cop.
No one, left or right, thinks there is street level corruption. Not the kind accessible to someone in a traffic stop. I have experienced it in Mexico and think that kind of corruption would still be worse because I cannot imagine how to recover from it. I have hope that a few high profile arrests of c level fall his may turn the tide. If not then there are extrajudicial methods open to American culture.
I know from my own personal experience that I haven't paid a bribe to a cop or to an alderman to get a zoning variance. There are some places where this kind of thing is routine. (e.g. I know there is a crooked cop somewhere but I also believe that if I tried to pay a bribe to a cop it wouldn't go well)
Thus I trust people's reports of street level corruption.
If it comes to perceptions of "corruption in high places" that is mediated by the media. It may well be that it is very corrupt and you never hear about it, or that it squeaky clean but you hear allegations 10 times a day. Or a Democrat might think everything is corrupt when Republicans are in power and then when Democrats are in power, Republicans take up the slack.
So I don't trust people's reports of corruption in high places.
Now I know a lot of people who are involved in road construction and maintenance in upstate NY who range from "drives a truck" to "manages $10M+ projects" and the belief that there is corruption in highway projects is widespread based on second- and third- hand accounts.
There's an interesting paradox hiding in plain sight here: Xi imported the anticorruption strat from Singapore, so superficially, PH below is correct.
Since SG has the opposite problem from China: ((low) gov corruption XOR (low) social trust).
The paradox is that the strat will have wildly different failure modes in SG vs CN or JP (all are aware of this!)
Note also that SG is consistently ranked in the top 10 together with the Nordics+NZ+IE+CH in spite of this failure mode..
My informed opinion is that LDP+loyalist-bureaucrats have been shipping and failing the same strat for years-- LDP dominance is the tell. Any critique of Xi's policy can be backed with JP data. Prediction: Xi will succeed if and only if CPC fall (or do the ship-of-theseus thing)
It's the institutional part which is lacking in France. Look at the budget of the ministry of justice in France per capita and in Germany. Germany spend twice as much and has twice as much judges per capita than France (and everything which goes with it like clerks).
My company took the biggest telecom company in France to court for a violation of our license on a soft, license was GPLv2, we won, but it took 12 years.
Justice is a very poor and slow institution in France. For the same countries the budget of police forces per capita are nearly the same for example.
Also Germany spends more than France on defence while having a lot less to show for, with France having nuclear weapons, nuclear subs, aircraft carriers and a much more capable military overall with less money. Germany is the poster child of government waste. If I were a taxpayer there I'd want my money back and/or bureaucrats going to jail.
> China's pretty corrupt politically but the social trust is quite high, the highest outside of northern europe as far as I can tell
There are a few reasons for that that I can imagine:
- China is one of very few autocracies that has managed to significantly improve the standard of living of most of its population.
- The public trials and (sometimes) executions of allegedly corrupt individuals might help improve the perception of corruption.
- The same harsh penalties mentioned above might influence people to declare a higher level of social trust than they actually have, even if the poll is supposedly "confidential" and "only for scientific purposes".
You can’t honestly say that a country where citizens inform on each other and put each other in forced labor camps based on rumor is a society where trust is high.
> The Deutscherblick ("German Look") was a tense, habitual glance over the shoulder used by citizens in Nazi-era Germany before speaking about sensitive topics like food rations, Hitler jokes, or the war’s progress
Not just labor camps. People would regularly get beheaded for anti-regime remarks. Nazi justice was keen on capital punishment for relatively minor crimes.
Nazi Germany was full of dedicated informers who would even earn money or other privileges for denouncing someone. It had about as much trust between strangers as Iran under the Revolutionary Guards might have today.
Now define "members". It's both possible and common for an in-group to experience a high degree of trust and care, while those outside that group to... Not. From the point of view of the beneficiaries the social contract is working beautifully!
I found Singapore somewhat bracing in how honestly they acknowledge the two tiers (natives + wealthy foreigners vs poor "guest" workers) in their society. The same division functionally exists in many "western" countries, but is broadly ignored. (To be clear, I do not endorse this - and, in fact, think it appalling - but appreciate straightforwardness more than I do obfuscation by empty rhetoric.)
It's not like atrocities started with Nazis. Child prostitution, high unemployment, corruption, poverty, moral devastation, drug addiction, injustice, inequality ... all that existed before Nazis so many people ignored the warnings that came with the Nazi party since they were the only ones promising to act.
Something about this doesn’t sit right with me. I’m not a historian but something tells me not everyone else was a-ok with atrocities that existed before nazis.
Also, again I’m not a historian, but I believe their promise to act was also tied up in blaming others and hate.
“At least they got things done” is very often the seed around which a belief in fascism crystallizes.
They don’t deserve any recognition for what they promised or what they accomplished.
This is a thread about my definition of a high trust society.
So far the only argument in support of nazi germany being high trust is that they got shit done.
I don’t see how anyone could argue that imprisoning your own population in forced labor camps based on rumor is something that can happen in a high trust society.
There is no trust in such a society, only fear.
Arguing about this any more is making me feel sick.
My only claim was that things were far from perfect (profoundly broken) before Nazis came to power. They made many terrible things, but they also fixed some of the issues they promised to fix. That's why they were able to grab power.
You can read more about the Weimar Republic. If it weren't so fundamentally broken, Nazis would never come to power. Stating this is not Nazi apologia but a warning of what happens if governments and the ruling class ignore the will of their own people and actively work against it for a long time.
The original context of this thread is the validity of china’s data in this trust survey. Interjecting with positive example of nazi germany is not even correct. There’s no way anyone can argue in good faith that nazi germany was high trust.
The nazis don’t need you to come to their defense.
Nazism is basically ignoring the will of their own people and actively working against them.
Nazis were one of the worst regimes we had in Europe, but the German regime before them was also bad. Bad enough that many saw Nazis as a viable alternative. In desperate times, people make desperate choices.
This is more or less a historical consensus, which I pointed out. If we want to prevent a Nazi-like regime from coming to power again, we need to avoid mistakes made by the Weimar Republic. Unfortunately, I see a global trend of governments making the same mistakes again, and I fear it will end in the same way.
I don't think Nazi society was high-trust. But I also don't think Nazis destroyed trust, because it already eroded before them.
China was getting better for a long time. XI is changing that. Change is slow though and he is not rushing corruption though it seems to be increasing. He has purged some corrupt people as well making things slightly better in the short term - but he values loyalty over competence and so his short term changes are for less corruption but long term increase it.
That is China is a complex country and books (which are not written and many cannot be for decades yet) are needed to understand this, not a short comment box. [This applies to every other country anyone here mentions]
Social trust is high because there are pretty heavy handed control measures over the population with havy costs. Thats more of a fear based society than trust. Government can giveth and government can taketh.
1. Fear of a capricious state can cause survival-motivated compliance which can appear as "trust" in coarse measurements. Meaning, you simply do fewer of those things that would provide opportunities for distrust in contexts where that could happen.
2. In a relatively severe, but consistent regime, the high penalties for violating trust in everyday cases (crime) act as a deterrent.
3. Fear may cause people to be selective and mindful about their social associations based on stronger proofs of trustworthiness. You might tell a Hitler joke to someone you have used more energy/caution to "vet", but avoid being too casual in environments of undetermined trustworthiness.
We are probably meant to assume ceteris paribus and only vary the dimension of corruption.
I think you’re right that culture plays a key role. For example if small bribes are customary, that doesn’t erode trust, that’s just the way things are.
Given that the AI scene is not stable at all in the current moment (every day a new release that make last month's obsolete), any offer that tries to lock you with a model or model provider is a bad idea.
Pay-per-use for the moment, until market consolidation and/or commoditization.
"While our genes are not changed by life experiences, they can be tuned through a system known as epigenetics."
It is indeed not a modification of the genetic code. And the transmission of epigenetic state from one generation to the next is much less straightforward.
But there is another lasting effect of the attack, hidden deep in the genes of
Syrian families. The grandchildren of women who were pregnant during the siege —
grandchildren who never experienced such violence themselves — nonetheless bear
marks of it in their genomes. Passed down through their mothers, this genetic
imprint offers the first human evidence of a phenomenon previously documented
only in animals: The genetic transmission of stress across multiple generations.
The article clearly implies a modify of the genes. The genome is altered.
The actual study doesn’t make that claim, the article is presenting it incorrectly. They’re talking about a methylation on certain genes. Think of it as amplification or attenuation and of a signal. The signal is the same, just weaker or stronger.
I've read the article, I've read the comments. Nobody has remotely tried to explain how these changes are heritable, at all. The article had an image with the word "germline", which was never expounded on.
The history of this tax makes me thing how SaaS prices calculated with some metrics that are imperfectly correlated with usage (per user, per host) leads to simple behaviour change to avoid an increase in pricing.
reply