Hacker Newsnew | past | comments | ask | show | jobs | submit | foobar2020's commentslogin

Chapter 2 has a plot of our understanding of the impact of all natural and human causes on the change in temperature: https://nca2018.globalchange.gov/img/figure/figure2_1.png

It's pretty clear (according to the models) that the human created aerosols combat a fair chunk of the increase caused by carbon emissions (around 0.7 deg C at the moment, which is substantial; we'd have already been around 2 deg C without it). Is deliberately increasing the amount of aerosols in the atmosphere a viable path forward? Is this economically feasible and safe? Can this be used to bring us back to the pre-industrial temperature levels?


There's lots of studies suggesting you can do this at low cost. There are issues, from not solving some problems at all (ocean acidification) to creating others (localized rainfall reduction).

My own thoughts are that we should be exploring doing this as a secondary backstop (it has pretty high ROI), while aggressively cutting emissions as well.

https://www.carbonbrief.org/solar-geoengineering-may-not-coo...


In my view, this could be dangerous if we do it before we are on the way to eliminating carbon via an alternative. If we get a symptom reduction via aerosols, we might just burn everything we can, and have some terrible effects on the oceans.

We might have to do it anyway, but I'm hoping we can hold off till we have a replacement, so as to avoid compounding things even worse than they are. Human psychology being what it is, we'll use aerosols to push off solving things.

Anyone know the expected outcome of a seriously acid ocean? My assumption is it would be civilization ending in some way, but maybe that's wrong.


It seems like we have to do some geoengineering no matter what, because the emission cuts required to solve the problem are unattainable even if every government on the planet was trying to do its part.

The current climate change problem has been caused by unintentional geoengineering with an undesirable outcome. It's reasonable to assume that different methods of geoengineering can have different effects, and might be capable of helping solve the problem.


Simple median is a very good (in many applications) nonlinear filter.


I don't know what it took but they did it recently:

https://news.ycombinator.com/item?id=17919297


> Poles and Italian immigrants in the US raised their average IQ scores over the 1st half of the 20th century -- by over 15 points!

So did everyone else: https://en.wikipedia.org/wiki/Flynn_effect

3 points per decade = 15 points per 50 years. So this shows that immigration to the US had no impact on the IQ of this group.

Whole developed world has very similar IQ levels to United States, except for China, Japan, Korea which have higher average levels: https://iq-research.info/en/average-iq-by-country


There are also a lot of very boring explanations for this. It turns out that good nutrition and avoiding lead poisoning (two major improvements in quality of life in developed countries in the 20th century) both lead to higher IQ.


Right right, I'm not sure anyone really knows why this happens.

My point is that 15 points per half-decade is just baseline. Doesn't mean we've picked up any useful correlation.


Doesn't mean we've picked up any useful correlation.

There are US sub-populations that have lagged. Genetics does not explain this, as overseas populations with the same genetics have not lagged.


Interesting. Do you have any sources where I could read about it?

Flynn effect is fascinating me - if it holds in the future we'll have, in developed nations, an average IQ of 130 (by today's standards) in 100 years. That's a LOT. It'd, roughly, mean that ~17% of people would be capable of graduating from today's Ivy League schools without problems (145 IQ or more). And there would still be the (new) top 0.5%. But whether it will hold is anyone's guess.


The Flynn Effect in Britain, Kenya, and Dominica:

Black IQ Gains in Britain, Kenya and Dominica

https://robertlindsay.wordpress.com/2011/01/19/black-iq-gain...

Eyeferth Study -- Children of Black American servicemen in Germany had the same average IQ as German children:

https://en.wikipedia.org/wiki/Eyferth_study

Thomas Sowell's survey of the work of Arthur Jensen and others. Basically, the patterns around the Flynn Effect are everywhere, and it shows up in Black populations as well. There is also a mention of a study of Black children raised by white families, which found the average IQ of that population to be 106.

https://www.nytimes.com/1977/03/27/archives/new-light-on-bla...


Flynn most recently has been giving talks about the reverse Flynn effect.

It isn't holding, especially in isolated populations that are easy to measure like Iceland.


Flynn effect stopped in the 1980s and even reversed since then.

Most probably it was due to the increasing ubiquity of K-12 education and fewer bad childrearing habits (smoking/drinking while pregnant, lead exposure, no vitamins).


> So did everyone else: https://en.wikipedia.org/wiki/Flynn_effect .

Then, look at differentials in the US that came after 1950.


It's 19% of the capital gains, not of the invested money. The parent comment makes it seem like the tax is absurd. Here's a list of countries that implement this tax, including Poland:

https://en.wikipedia.org/wiki/Capital_gains_tax

You _can_ save and invest in Poland.


Right, I might phrased it imprecisely. Anyway let me show you screenshot from my bank taken today (sorry it's for some reason only in polish): https://imgur.com/a/ISthS

so putting 10k PLN (2k EUR) for 24 months in, you'll gain 50 EUR and small text at the bottom says it's BEFORE taxation! Average gross wage depends on region, but it's around 1k EUR per month. I cannot find median, values but it's much lower, most people here don't have two cents to rub together not to mention to put twice the average salary for 24 months to get 50 EUR


Well there are more people that are willing to pay than just Coinbase. A more comprehensive answer would be that arbitrage is not that up to speed in the Bitcoin world, especially in KRW and JPY markets: https://coinmarketcap.com/currencies/bitcoin/#markets

Used to be that crypto is massively overpriced there, now it looks like Korea and Japan are leading this price fall.


> A more comprehensive answer would be that arbitrage is not that up to speed in the Bitcoin world...

Or that folks trust their ability to get money out of Coinbase more than the other exchanges.


Unicode to the rescue! Here's how to make 𝗯𝗼𝗹𝗱 text:

http://qaz.wtf/u/convert.cgi

Looks a bit ugly, but works.


This would be incredibly useful for distributed machine learning - imagine a Tensorflow implementation that almost entirely bypasses CPU.


For most applications, getting training images onto the GPU isn't the bottleneck by far. Training the Inception model, for example, handles batches of 32 images (299x299x3) in 1.2 seconds. That's a pretty boring ~300KB * 32 ~= 10MB/sec of read bandwidth off the SSD for Imagenet. Even dealing with "real" images is probably only 10x that, which is trivial to get over the PCI bus.

The question would be whether we can turn the crank on the design of models to make it possible to do something really cool given access to very high-speed SSD storage.


I was thinking the same thing, but is SSD to GPU faster than RAM to GPU? In many (not all) cases you buy a tonne of RAM and load your entire dataset into memory once and then iterate over it as necessary.

You also lose the flexibility of doing any sort of data modification or augmentation. One domain where your data usually doesn't fit in RAM is image recognition, but often you want to do things like apply random flips, crops and change hues before training to make the neural net less sensitive to those changes, which you can't really do with this.


SSD is probably not as fast as RAM, but it's much much cheaper, in the order of 10x per gigabyte. With SSD-GPU bridge you can have fast access to a multiple TiB training set, on a single machine.

Data pre-processing is indeed an issue, but hue adjustment/flipping/cropping could be implemented as Tensorflow operations, on the GPU. Similarly with input decompression - it would either have to be done on GPU, or the data would have to be stored uncompressed.


As long as the average bandwidth isn't a bottleneck, it's not going to matter - at worst, you're just going to need to prefetch (and due to SSD latency, that's likely optimal regardless).


RAM-to-GPU is always faster than SSD-to-GPU. It is a solution to help a situation when data size does not fit RAM size (or when user has less budget to purchase enough RAM. In fact, we can purchase Intel SSD 750 (400GB) with 300USD).


For the scenario you're targeting: databases, this makes a tonne of sense, database data regularly exceeds the size of RAM and the operations you want to do on the data are pretty static in the sense that they're the SQL operators.

In deep learning you are usually doing a lot more custom processing and your datasets are usually not as big, such that just buying more RAM is often cost effective.


They do, if though nothing else then through multi-billion acquisitions. Another thing is that we all know crises come every now and then. Being tight on cash when it hits the market means you'll have to let talent go to your competitors, slow down infrastructure growth, reduce research freedom.


Sure, these are reasonable arguments, which are commonly used. Clearly this justifies holding on to some cash. But I question whether it justifies holding on to this much cash. At some point it has to reach diminishing returns.


The problem with spending it on acquisitions is once you reach a certain size buying other companies becomes difficult due to regulatory concerns.


Without disagreeing that some portion of global profit comes from "taking things from the planet", do you think that all the value we've created since emerging from Africa somewhen between 50.000 and 100.000 years ago can be compensated with environmental damage? Or is it a recent trend? If we were to plot a graph, where the X axis is the time and the Y axis is profits/damage, how would it look like?

What about mathematics, arts and science? It sure has incredible value and eventually translates into corporate profit. For the most part I cannot see the corresponding environmental debt.

And to be more specific: what kind of damage did the three mentioned companies do?


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: