But a) that's the cost to the user -- we don't know how much loss they're taking on those and b) the number of tokens to serve a similar prompt has been going up, so that the total cost to serve a prompt has been going up in general. Any cost analysis that doesn't mention these is hugely misleading.
I don't think that's right, even for laypeople. It's just that the pain of things that take 5 seconds when they could take 50 ms is subtle and can be discounted or ignored until you are doing a hundred things in a row that take 5 seconds instead of 50 ms. And if you don't know that it should be doable in 50 ms then you don't necessarily know you should be complaining about that pain.
It's also that the people who pay the price for slowness aren't the people who can fix it. Optimizing a common function in popular code might collectively save centuries of time, but unless that converts to making more money for your employer, they probably don't want you to do it. https://www.folklore.org/Saving_Lives.html
> [...] three factors [...] Capital accumulation is one.
The obvious omission here is well developed in Imperialism, The Highest Stage of Capitalism: it's hard to accumulate capital when all of the productivity growth from foreign "investment" by the rich world is captured by the "investors".
A huge fraction of that assistance (e.g. the IMF) has been conditioned on opening up their economies to "fair" foreign exploitation. It's not some benevolent gift, generally.
It's extremely naive to ignore the power dynamics that set the terms of who benefits from those investments, and the relationship between those power dynamics and the initial wealth of the participants. I'll withhold judgement on whether it's "cope".
Edit to add: point being that a weak correlation, even if it is indisputably real, leaves a lot of room for other factors to be operative when it comes to particular differences.
rank correlation between GDP and life expectancy (child mortality would have been maybe a bit better). uses a 20 year window in both directions per year. bootstraps for 5% and 95% quantiles of the rank correlation.
there looks to be a max around 1992, and steady downhill in correlation since then.
this seems unlikely to be an artifact of the analysis, though 1992 is eerily close to 20 years from the last date. 2013 is the last year where we have at least 15 years total symmetriclaly around the given year to include in the correlation for that year.
I notice that the two most obvious outliers for low GDP high life expectancy countries are North Korea and Syria, and the low GDP low life expectancy outliers are three African petrostates.
My interpretation is that this confirms both points. Yes, petrostates with concentrated wealth, states with dubious truthfulness, and those ravaged by war are all cases where GDP doesn't tell the full story. I'm not sure that tells us GDP is useless when applied to where most HN users live, though.
I didn't say it was useless, just that saying there's a correlation is not a refutation of OP's claim that financialization is an important confounding factor.
More to the point, it looks like you are attempting to identify some of the particular factors that might affect gdp or health outcomes in ways that aren't correlated, and exclude cases where those are the operative factors. Which would provide support for the thesis that there are other operative factors, as suggested by OP.
It’s not cherry-picking. The question is: does GDP correlate with things we care about? In answering that question it makes sense to hold other factors constant. Asians have a longevity advantage in virtually every society where they’re found, so it makes sense to separate out the asian countries. It also makes sense to remove outliers where a single export makes GDP per capita seem artificially large.
1. The blog states that the relationship between economic growth and initial wealth doesn't show a trend that had been expected from theory. It uses per capita GDP and per capita GDP growth as proxies for wealth and growth. Its analysis does not control for any pet theories, because the theory claims to be general. (Indeed, it refers to literature that does control for other factors and explains why they are unsatisfying: because if the other factors are more important than the economic ones, then the theory isn't worth much)
2. OP, roughly, objects that GDP (and its growth rate) is sensitive to a fairly specific fact of the modern global economy, namely uneven financialization, and hints at an argument for why GDP would obscure the relevant trends in wealth and growth of wealth. They claim that the dynamics of various health proxies are better representative of economic status and growth than GDP is. They do not claim that these are not correlated. Effectively, their claim is that the portion of the variation of GDP that is not explained by its correlation with health outcomes is enough to obscure the relationship between economic status and growth.
3. You comment tersely that they are correlated, which was never denied in OP, and does not address in any way the relevance of the particular confounding factor that OP proposed. In followups try to rescue that correlation by controlling for your pet "extra factors" that (1) objected to for how they weaken the relevance of the economic theory, and that are irrelevant to the point that OP was trying to make.
To be clear, I have no strong evidence that financialization explains the way that GDP (and its derivative) differs from a linear function of other development metrics (and their derivatives), and if OP has evidence, they don't provide it in their brief HN comment. All I'm saying is, saying there's a correlation has very little bearing on OP's claim, and controlling for more irrelevant pet factors is not going to make it more relevant.
> The question is: does GDP correlate with things we care about?
No that is not the question. The questions are: a) would controlling for financialization be a better test of the economic theory, and b) are health metrics a good way of controlling for the portion of GDP that is explained by GDP.
Again, I don't know the answers, just that your cherry-picked pet factors are irrelevant to these questions.
The writer presumably knows that umlauts and other non-ascii characters are functional in many languages. "rock döts" is poking fun at the trend in a certain tranche of anglophone rock/metal to use them in a purely aesthetic way in band names etc.
In which case doing this in the dark is clearly bad for the community -- if that location is what's scarce then they should be demanding a better deal.
The only way this makes sense for communities is as a kind of "finder's fee", i.e. you might argue that if BigTechOne™ knew that they'd have to bid against BigTechTwo™ they'd never even bother to scope out the location.
Still, if the prospecting is the bottleneck there could be 3rd parties (or even the tech companies themselves) entering into agreements with towns which allow both a finders fee and open bidding for the lot.
reply